diff --git "a/querist.csv" "b/querist.csv" --- "a/querist.csv" +++ "b/querist.csv" @@ -3,7 +3,7 @@ AIBD-0789,2022-01-01,event,month,reported,"FDA finds that the Jewel patch defibr AIBD-1141,2024-01-01,event,year,reported,"Austin's City Auditor finds that APD's license plate readers led to 165 arrests from March to December 2024, with about 133 prosecutions",Austin's City Auditor found that APD's license plate readers led to 165 arrests from March to December 2024. About 133 prosecutions followed. The readers helped recover a missing person.,Austin's *license plate readers* led to 165 arrests in a 2024 trial,"165 arrests, about 133 prosecutions",,found,Found by an auditor,no comparison,very low,aggregate result,false,Automated License Plate Reader (ALPR) program with Flock Safety fixed cameras and Axon vehicle-mounted cameras,,Flock Safety (fixed cameras and ALPR service) and Axon Enterprise (vehicle-mounted cameras),United States,Austin Police Department (APD),United States,"Austin, Texas, United States",United States,Policing,,Safety and rescue,Images and video,Public servants,16,High income,physical_health,Crime reduction or prevention,4,1,3,,3,5,,organizational; personal; societal,1,0,,,,,withdrawn,2025-06-30,shown,https://builtin.com/articles/flock-cameras,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Audit report"", ""value"": ""City of Austin Office of the City Auditor. APD License Plate Reader. Audit Report, May 2025. Finding 1, Resolution item C and Exhibit 3, pp. 6-7. The auditor analyzed Flock ALPR data and APD Risk Management's quarterly audits for March to December 2024."", ""url"": ""https://services.austintexas.gov/edims/document.cfm?id=452710""}, {""label"": ""Audit report"", ""value"": ""City of Austin Office of the City Auditor. APD License Plate Reader. Audit Report, May 2025. Finding 1, Resolution item B, p. 6. The auditor reviewed all 286 Hit Stop Forms from the three quarters."", ""url"": ""https://services.austintexas.gov/edims/document.cfm?id=452710""}, {""label"": ""Audit report"", ""value"": ""City of Austin Office of the City Auditor. APD License Plate Reader. Audit Report, May 2025. Finding 1, Resolution item A, p. 5. Because of these gaps, the auditor could not verify the number of Flock scans."", ""url"": ""https://services.austintexas.gov/edims/document.cfm?id=452710""}, {""label"": ""Follow-up"", ""value"": ""Austin's city manager withdrew the plate reader program from the council agenda in June 2025, and APD's program effectively ended on 30 June 2025."", ""url"": ""https://austinmonitor.com/stories/2025/06/council-divided-on-reader-program/""}]","{""kind"": ""Withdrawals""}" AIBD-2312,2021-01-01,source publication,month,reported,"St George's Hospital, London: preterm preeclampsia fell from 0.84% to 0.56% of pregnancies after the hospital began to screen with the Fetal Medicine Foundation's algorithm, and Ontario Health judged that such programs likely reduce the risk","At St George's Hospital in London, preterm preeclampsia fell from 0.84% to 0.56% of pregnancies after the hospital began to screen with an algorithm, and Ontario Health judged the reduction likely.",*FMF screening* cut preterm preeclampsia at a London hospital,0.84% to 0.56%,,found,Found by an assessment body,before and after,very low,aggregate result,false,"Fetal Medicine Foundation (FMF) first-trimester preeclampsia screening algorithm (maternal factors, blood pressure, PAPP-A and uterine artery Doppler)",,Fetal Medicine Foundation,United Kingdom,St George's University Hospitals NHS Foundation Trust,United Kingdom,"London, United Kingdom",United Kingdom,Maternity care (preeclampsia screening),,Other,Prediction from data,Women and mothers,3,High income,physical_health,Medical diagnosis,4,3,3,,3,5,,personal,3,0,Women screened at the same hospital under national guidance before the change,0.84% had preeclampsia with delivery before 37 weeks,0.56%,measured,unknown,,shown,https://pubmed.ncbi.nlm.nih.gov/26297382/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Ontario Health. First-Trimester Screening Program for the Risk of Pre-eclampsia Using a Multiple-Marker Algorithm: A Health Technology Assessment. Ontario Health Technology Assessment Series 22(5):1-118, December 2022. Ontario Health is the agency of the Government of Ontario that assesses health technologies for public funding."", ""url"": ""https://pmc.ncbi.nlm.nih.gov/articles/PMC10530459/""}, {""label"": ""Assessment report"", ""value"": ""Ontario Health. First-Trimester Screening Program for the Risk of Pre-eclampsia Using a Multiple-Marker Algorithm: A Health Technology Assessment. Ontario Health Technology Assessment Series 22(5):1-118, December 2022. Ontario Health is the agency of the Government of Ontario that assesses health technologies for public funding."", ""url"": ""https://pmc.ncbi.nlm.nih.gov/articles/PMC10530459/""}, {""label"": ""Assessment report"", ""value"": ""Ontario Health. First-Trimester Screening Program for the Risk of Pre-eclampsia Using a Multiple-Marker Algorithm: A Health Technology Assessment. Ontario Health Technology Assessment Series 22(5):1-118, December 2022. Ontario Health is the agency of the Government of Ontario that assesses health technologies for public funding."", ""url"": ""https://pmc.ncbi.nlm.nih.gov/articles/PMC10530459/""}, {""label"": ""Assessment report"", ""value"": ""Ontario Health. First-Trimester Screening Program for the Risk of Pre-eclampsia Using a Multiple-Marker Algorithm: A Health Technology Assessment. Ontario Health Technology Assessment Series 22(5):1-118, December 2022. Ontario Health is the agency of the Government of Ontario that assesses health technologies for public funding."", ""url"": ""https://pmc.ncbi.nlm.nih.gov/articles/PMC10530459/""}, {""label"": ""Assessment report"", ""value"": ""Ontario Health. First-Trimester Screening Program for the Risk of Pre-eclampsia Using a Multiple-Marker Algorithm: A Health Technology Assessment. Ontario Health Technology Assessment Series 22(5):1-118, December 2022. Ontario Health is the agency of the Government of Ontario that assesses health technologies for public funding."", ""url"": ""https://pmc.ncbi.nlm.nih.gov/articles/PMC10530459/""}]","{""kind"": ""Official documents""}" AIBD-2237,2019-02-08,source publication,day,mixed,"ODI Resume, Investigation PE 16-007: Automatic vehicle control systems",The U.S. road safety regulator stated in 2017 that Tesla's rate of airbag-deployment crashes fell by almost 40% after Autosteer. A 2019 reanalysis of the data found a 59% rise.,NHTSA's finding of 40% fewer *Autosteer* crashes is later challenged,fell by almost 40%,,found,Found by a regulator,before and after,very low,harm prevented,false,Tesla Autosteer (part of Autopilot),,Tesla,United States,Owners of Tesla Model S and Model X vehicles of model years 2014 to 2016,United States,United States,United States,Road vehicle crash avoidance,,Safety and rescue,Machines that act,,3; 11,High income,physical_health,Hazard reduction,4,4,,,2,,,personal,2,0,The same vehicles before Autosteer was installed,The airbag-deployment crash rate before installation,A rate almost 40% lower after installation,measured,challenged,2019-02-08,shown,https://web.archive.org/web/20151016033541/https://www.teslamotors.com/blog/your-autopilot-has-arrived,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Regulatory document"", ""value"": ""National Highway Traffic Safety Administration, Office of Defects Investigation. Closing report of investigation PE 16-007 into Tesla's Autopilot and automatic emergency braking, closed 19 January 2017. The agency computed the crash rates itself from mileage and airbag data that Tesla supplied."", ""url"": ""https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF""}, {""label"": ""Regulatory document"", ""value"": ""National Highway Traffic Safety Administration, Office of Defects Investigation. Closing report of investigation PE 16-007 into Tesla's Autopilot and automatic emergency braking, closed 19 January 2017. The agency computed the crash rates itself from mileage and airbag data that Tesla supplied."", ""url"": ""https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF""}, {""label"": ""Follow-up"", ""value"": ""Quality Control Systems reanalyzed the agency's data in February 2019 and found that the crash rate rose by 59% among the vehicles with complete mileage data."", ""url"": ""https://www.safetyresearch.net/Library/NHTSA_Autosteer_Safety_Claim.pdf""}]","{""kind"": ""Challenges""}" -AIBD-0788,2026-03-03,source publication,day,reported,FDA finds that the Claire OCT System's AI helped surgeons clear residual breast cancer at lumpectomy margins in 7 of 35 patients,"In a trial that FDA reviewed, surgeons using the Claire OCT System after standard care cleared residual cancer in 7 of 35 patients, a 20.0% relative reduction.",*Claire OCT System* helped surgeons remove cancer left at the margins,a 20.0% relative reduction,,found,Found by a regulator,,very low,aggregate result,false,Claire OCT System with the ImgAssist AI detection algorithm,ImgAssist 2.0,"Perimeter Medical Imaging AI, Inc.",Canada,Breast surgeons at the investigational sites of the randomized trial (NCT05113927),United States,"Investigational sites, United States",United States,Healthcare,,Health,Images and video,Health workers; Patients,3,High income,physical_health,Medical diagnosis,3,2,1,,3,1,,personal; organizational,2,0,The same patients' margins after standard care alone,17.0% with at least one unaddressed positive margin,13.6% after use of the device,measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf25/P250008B.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. Summary of Safety and Effectiveness Data (SSED), P250008: Claire OCT System. Approved 2026-03-03."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf25/P250008B.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0788,2026-03-03,source publication,day,reported,FDA finds that the Claire OCT System's AI helped surgeons clear residual breast cancer at lumpectomy margins in 7 of 35 patients,"In a trial that FDA reviewed, surgeons using the Claire OCT System after standard care cleared residual cancer in 7 of 35 patients, a 20.0% relative reduction.",*Claire OCT System* helped surgeons remove cancer left at the margins,a 20.0% relative reduction,,found,Found by a regulator,controlled comparison,low,aggregate result,false,Claire OCT System with the ImgAssist AI detection algorithm,ImgAssist 2.0,"Perimeter Medical Imaging AI, Inc.",Canada,Breast surgeons at the investigational sites of the randomized trial (NCT05113927),United States,"Investigational sites, United States",United States,Healthcare,,Health,Images and video,Health workers; Patients,3,High income,physical_health,Medical diagnosis,3,2,1,,3,1,,personal; organizational,2,0,The same patients' margins after standard care alone,17.0% with at least one unaddressed positive margin,13.6% after use of the device,measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf25/P250008B.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. Summary of Safety and Effectiveness Data (SSED), P250008: Claire OCT System. Approved 2026-03-03."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf25/P250008B.pdf""}]","{""kind"": ""Official documents""}" AIBD-0193,2020-02-20,event,day,reported,"Data from eLocust3m, cleaned by AI, supports FAO's control of locusts in East Africa, which may have averted USD 1.56 billion in losses","PlantVillage's eLocust3m app gathered over 240,000 locust records. AI removed unusable ones for the control operation that FAO credits with averting USD 1.56 billion in losses.",Locust control using eLocust3m data may have averted USD 1.56 billion,protected the food security of 36.6 million people and avoided USD 1.56 billion in losses,,found,Found by an evaluation office,no comparison,very low,harm prevented,false,eLocust3m (PlantVillage) with AI data validation feeding FAO's Desert Locust Information Service,eLocust3M,"PlantVillage, Pennsylvania State University, with FAO",United States,"FAO Desert Locust Information Service and national locust control operations in Kenya, Ethiopia and Somalia",Italy,,,Agriculture / pest early warning,Desert locust control,Agriculture and food,Images and video,Public servants; People at risk of disaster,2,,environmental_impact; operational_improvement,Environmental impacts,3,5,4,4,3,5,0.640,societal; organizational; personal,7,2,no locust control operations in East Africa (FAO's estimate of the losses that the control campaign averted),"loss of 4 million tonnes of cereal and 800 million litres of milk production, USD 1.56 billion in cereal and milk losses",those losses prevented and the food security of 36.6 million people protected,assumed,persisted,2022-06-15,shown,https://www.fao.org/newsroom/story/A-high-tech-response-is-helping-countries-win-the-battle-against-Desert-Locusts/en,USD 1.56 billion,cereal and milk losses avoided by the locust control operations in East Africa that the app and AI data cleaning fed,True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""FAO Office of Evaluation (OED) (2020). Real-time evaluation of FAO's response to desert locust upsurge 2020-2021 - Phase I, p. 16 (PDF p. 20), finding 3.5.3, paras 44-46."", ""url"": ""https://www.fao.org/3/cb2437en/cb2437en.pdf""}, {""label"": ""Evaluation report"", ""value"": ""FAO Office of Evaluation (2021). Real-time evaluation of FAO's response to the desert locust upsurge 2020-2021 - Phase II. Programme Evaluation Series 10/2021. Rome: FAO (last updated 23 February 2022). Findings, paragraph 52 (confidence rating: very high); paragraph 25 on Kenya's 3,000 scouts trained on eLocust3g and eLocust3m."", ""url"": ""https://www.fao.org/3/cb6995en/cb6995en.pdf""}]","{""kind"": ""Official documents""}" AIBD-0717,2026-10-31,source access,day,reported,Evidence for ESSA rates DreamBox Math Strong on two randomized studies with math gains for students in kindergarten to grade 2,"Evidence for ESSA rates DreamBox Math Strong. In two randomized studies, users in kindergarten to grade 2 outperformed controls, while gains on state tests in grades 4–5 were not meaningful.",*DreamBox Math* earns a Strong rating for early math gains,the effect size is +0.10,,found,Found by an assessment body,randomized trial,high,aggregate result,false,DreamBox Math (adaptive online mathematics program),,DreamBox Learning,United States,Schools in the two studies: three high-poverty charter schools in the San Francisco Bay Area and 24 schools across North Carolina,United States,"California and North Carolina, United States",United States,Education,,Education,,Students; Children,4,High income,incentivization,Increased/personalized knowledge,2,,3,4,3,,,personal,1,0,Control students in the North Carolina randomized study,,"effect size = +0.12 (K-2, Number Knowledge Test), +0.03 (grades 4-5, end-of-grade test)",measured,persisted,2025-01-23,not shown,https://www.evidenceforessa.org/program/dreambox-learning/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Evidence for ESSA, Center for Research and Reform in Education, Johns Hopkins University. DreamBox Math: program page, rating Strong. Accessed 31 October 2026."", ""url"": ""https://www.evidenceforessa.org/program/dreambox-learning/""}]","{""kind"": ""Official documents""}" AIBD-0720,2026-06-01,source publication,day,reported,The What Works Clearinghouse finds promising evidence that Project LISTEN's Reading Tutor raised literacy scores in a randomized trial,The What Works Clearinghouse found promising (Tier 3) evidence for Project LISTEN's Reading Tutor. Students in grades 1 to 4 who read aloud to it gained more in literacy than peers in silent reading.,A listening *Reading Tutor* beats silent reading in a trial,improvement index 27,,found,Found by an assessment body,randomized trial,high,aggregate result,false,Project LISTEN's Reading Tutor (speech-recognition reading tutor),2000-2001 version,"Project LISTEN, Carnegie Mellon University",United States,Two elementary schools in an affluent suburban school district (not named),United States,"United States (suburban school district, not named)",United States,Education,,Education,Speech and sound,Children; Students,4,High income,incentivization,Increased/personalized knowledge,2,2,3,2,3,,,personal,3,0,Sustained silent reading in class during the same daily period,"471.49 (comparison group mean, Woodcock Reading Mastery Test total reading composite)",480.01 (Reading Tutor group mean),measured,scaled,2024-07-24,shown,https://www.cs.cmu.edu/~listen/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""U.S. Department of Education, Institute of Education Sciences, What Works Clearinghouse. Single study review of Mostow, Nelson-Taylor and Beck (2013), Computer-Guided Oral Reading versus Independent Practice. Reviewed June 2026."", ""url"": ""https://ies.ed.gov/ncee/WWC/Study/95240""}]","{""kind"": ""Official documents""}" @@ -26,7 +26,7 @@ AIBD-1009,2024-07-01,event,month,reported,ANAO finds that an AI-enabled model he AIBD-0747,2024-06-05,source publication,day,mixed,The Peterson Health Technology Institute finds that app-based exercise therapy such as Kaia improves pain compared with no physical therapy,PHTI found that app-based exercise therapy improves pain compared with no physical therapy. It rests on two German Kaia studies. PHTI rated the category no substitute for in-person therapy.,*Kaia*'s exercise app eased back pain in an independent assessment,,,found,Found by an assessment body,controlled comparison,low,aggregate result,false,"Kaia app for app-based exercise therapy, rated with DarioHealth's program in the same category",,Kaia Health,United States,"Kaia Health, through its app, in two studies of people with back pain in Germany",Germany,"Germany (studies), with an assessment for US purchasers",Germany | United States,Healthcare,,Health,Images and video,Patients,3,High income,physical_health,Medical access,2,3,3,,2,,,personal,2,0,Standard care for low back pain under German guidelines,,Pain improvements that met the MCID threshold in the app group only,measured,persisted,2026-01-28,shown,https://www.phti.org/wp-content/uploads/2024/10/PHTI-Virtual-MSK-Solutions-Assessment-Report-v1.3.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Peterson Health Technology Institute. Virtual Musculoskeletal Solutions. Health technology assessment, executive summary, 5 June 2024."", ""url"": ""https://www.phti.org/wp-content/uploads/2024/06/PHTI-Virtual-MSK-Solutions-Executive-Summary-1.pdf""}, {""label"": ""Assessment report"", ""value"": ""Peterson Health Technology Institute. Virtual Musculoskeletal Solutions: Assessment Report, version 1.3, updated 10 October 2024, technology context."", ""url"": ""https://www.phti.org/wp-content/uploads/2024/10/PHTI-Virtual-MSK-Solutions-Assessment-Report-v1.3.pdf""}]","{""kind"": ""Official documents""}" AIBD-0232,2024-05-17,event,day,reported,"Cheshire Police facial recognition matches CCTV to an 18-year-old county lines dealer, who gets 28 months while a second dealer gets 34",Cheshire Constabulary's facial recognition system matched CCTV of an unknown Northwich dealer to an 18-year-old. Police stopped him with a 19-year-old on 17 May 2024. Both were jailed on 1 July 2024.,Two men are jailed after face recognition identifies a Northwich dealer,jailed the 18-year-old for 28 months and the 19-year-old for 34 months,,found,Found by an auditor,no comparison,very low,benefit created,false,Cheshire Constabulary facial recognition system (retrospective),,Not named in the sources,,"Cheshire Constabulary, Northwich Proactive Policing Unit",United Kingdom,"Northwich, United Kingdom",United Kingdom,Policing / drug supply,Government,Public sector,Images and video,Public servants,16,High income,physical_health,Crime reduction or prevention,2,,4,,4,5,,societal,4,1,conventional identification of a suspect who was not known locally,The 18-year-old was not known in the Northwich area,we were quickly able to identify him as a suspect,assumed,scaled,2026-02-10,not shown,https://assets-hmicfrs.justiceinspectorates.gov.uk/uploads/peel-assessment-2023-25-cheshire.pdf,£990,"crack cocaine seized from the 18-year-old (99 wraps, approximate street value)",False,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""His Majesty's Inspectorate of Constabulary and Fire & Rescue Services. PEEL 2023-2025: An inspection of Cheshire Constabulary. Published 6 February 2025. Section 'Preventing and deterring crime and antisocial behaviour, and reducing vulnerability', innovative practice 'Cheshire Constabulary uses retrospective facial recognition to effectively identify suspects and tackle criminality'. The case counts are the force's own figures, and the inspection does not examine this case."", ""url"": ""https://assets-hmicfrs.justiceinspectorates.gov.uk/uploads/peel-assessment-2023-25-cheshire.pdf""}]","{""kind"": ""Official documents""}" AIBD-0742,2024-04-01,source publication,month,reported,Health Technology Wales supports routine use of AI polyp detection in colonoscopy after estimating that it raised adenoma detection by 23%,Health Technology Wales pooled 36 randomized trials. It estimated that AI-assisted colonoscopy found adenomas in 44.3% of people against 37.1% without it. It supports routine adoption.,Wales supports *CADe* colonoscopy after it found more adenomas in trials,23% higher with CADe than with standard colonoscopy,,found,Found by an assessment body,randomized trial,low,aggregate result,false,"Computer-aided detection (CADe) colonoscopy systems in the trials that HTW pooled, including GI Genius, ENDO-AID, CAD EYE, EndoScreener and EndoAngel",,"Medtronic (GI Genius), Olympus (ENDO-AID), Fujifilm (CAD EYE), Wision AI (EndoScreener), Wuhan EndoAngel Medical Technology Company (EndoAngel) and other makers","Multiple: Ireland, Japan, China",Endoscopy services in the pooled randomized trials,Multiple,"Multiple countries (trials), with guidance for Wales, United Kingdom",United Kingdom,Healthcare,,Health,Images and video,Health workers,3,High income,physical_health,Medical diagnosis,2,4,1,,3,1,,personal,3,0,Standard colonoscopy without CADe in the pooled randomized trials,37.1% (standard colonoscopy),"44.3% (CADe), RR: 1.23",measured,unknown,,shown,https://healthtechnology.wales/wp-content/uploads/EAR055-AIAE-WEB.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Health Technology Wales. Artificial Intelligence (AI)-assisted endoscopy in the detection of gastrointestinal cancer and pre-cancerous lesions. Guidance GUI055, April 2024 (topic page, outcome of the appraisal)."", ""url"": ""https://healthtechnology.wales/reports-guidance/ai-assisted-endoscopy-for-gastrointestinal-cancer/""}, {""label"": ""Assessment report"", ""value"": ""Health Technology Wales. Evidence Appraisal Report EAR055: Artificial intelligence assisted endoscopy in the detection of lower gastrointestinal cancer and pre-cancerous lesions. April 2024, section 5.2.1 and Table 2."", ""url"": ""https://healthtechnology.wales/wp-content/uploads/EAR055-AIAE-WEB.pdf""}, {""label"": ""Assessment report"", ""value"": ""Behandlingsrådet (Danish Health Technology Council). Kunstig intelligens som beslutningsstøtte ved koloskopiske undersøgelser til diagnosticering af neoplastisk sygdom. Recommendation in force from 1 February 2023. Archived copy of 15 September 2025 of the council's page. The quotations are in Danish."", ""url"": ""https://web.archive.org/web/20250915185757id_/https://behandlingsraadet.dk/find-evalueringer-analyser/analyser/2021/kunstig-intelligens-som-beslutningsstotte-ved-koloskopiske-undersogelser-til-diagnosticering-af-neoplastisk-sygdom""}, {""label"": ""Assessment report"", ""value"": ""Behandlingsrådet (Danish Health Technology Council). Kunstig intelligens som beslutningsstøtte ved koloskopiske undersøgelser til diagnosticering af neoplastisk sygdom. Recommendation in force from 1 February 2023. Archived copy of 15 September 2025 of the council's page. The quotations are in Danish."", ""url"": ""https://web.archive.org/web/20250915185757id_/https://behandlingsraadet.dk/find-evalueringer-analyser/analyser/2021/kunstig-intelligens-som-beslutningsstotte-ved-koloskopiske-undersogelser-til-diagnosticering-af-neoplastisk-sygdom""}]","{""kind"": ""Official documents""}" -AIBD-1161,2024-03-28,source publication,day,reported,"Japan's health ministry makes EndoBRAIN-EYE eligible for an insurance add-on, recognizing that it raises the accuracy of specialists in colonoscopy","Japan's health ministry recognized EndoBRAIN-EYE as shown to raise specialists' diagnostic accuracy in colonoscopy, which made it eligible for a new insurance add-on to polyp removal.",*EndoBRAIN-EYE* wins a Japanese insurance add-on for better colonoscopy,tumor detection rate from 21.6% to 31.2%,,found,Found by an assessment body,,very low,aggregate result,false,EndoBRAIN-EYE (内視鏡画像診断支援プログラム),,"Cybernet Systems Co., Ltd. (サイバネットシステム株式会社), with Showa University Northern Yokohama Hospital and Nagoya University",Japan,Japanese hospitals and clinics where colonoscopy specialists use it during polyp removal (not named),Japan,Japan,Japan,Healthcare,,Health,Images and video,Health workers,3,High income,physical_health,Medical diagnosis,2,,1,,3,1,,personal,3,0,Colonoscopy by specialists without the program,21.6% tumor detection rate,31.2% tumor detection rate,measured,unknown,,shown,https://www.cybernet.co.jp/medical-imaging/products/endobrain/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Ministry of Health, Labour and Welfare (MHLW), Health Insurance Bureau, Medical Economics Division, Japan. 疑義解釈資料の送付について(その1)(Q&A on the interpretation of the fiscal 2024 medical fee schedule, part 1), office communication of 28 March 2024. Attachment 1 (medical fee schedule), Question 221, p. 医-58."", ""url"": ""https://www.mhlw.go.jp/content/12404000/001237675.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Pharmaceuticals and Medical Devices Agency (PMDA), Japan. 資料2, PMDA提出資料, 第22回 医療機器・体外診断薬の承認審査や安全対策等に関する定期意見交換会 (Document 2, submitted by PMDA to the 22nd regular exchange meeting on the review and safety of medical devices and in vitro diagnostics), 7 October 2024. Table AIを活用した医療機器(プログラム)の承認状況(R6.9末現在) (approval status of AI-based program devices as of the end of September 2024)."", ""url"": ""https://www.mhlw.go.jp/content/11124500/001329696.pdf""}]","{""kind"": ""Official documents""}" +AIBD-1161,2024-03-28,source publication,day,reported,"Japan's health ministry makes EndoBRAIN-EYE eligible for an insurance add-on, recognizing that it raises the accuracy of specialists in colonoscopy","Japan's health ministry recognized EndoBRAIN-EYE as shown to raise specialists' diagnostic accuracy in colonoscopy, which made it eligible for a new insurance add-on to polyp removal.",*EndoBRAIN-EYE* wins a Japanese insurance add-on for better colonoscopy,tumor detection rate from 21.6% to 31.2%,,found,Found by an assessment body,before and after,very low,aggregate result,false,EndoBRAIN-EYE (内視鏡画像診断支援プログラム),,"Cybernet Systems Co., Ltd. (サイバネットシステム株式会社), with Showa University Northern Yokohama Hospital and Nagoya University",Japan,Japanese hospitals and clinics where colonoscopy specialists use it during polyp removal (not named),Japan,Japan,Japan,Healthcare,,Health,Images and video,Health workers,3,High income,physical_health,Medical diagnosis,2,,1,,3,1,,personal,3,0,Colonoscopy by specialists without the program,21.6% tumor detection rate,31.2% tumor detection rate,measured,unknown,,shown,https://www.cybernet.co.jp/medical-imaging/products/endobrain/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Ministry of Health, Labour and Welfare (MHLW), Health Insurance Bureau, Medical Economics Division, Japan. 疑義解釈資料の送付について(その1)(Q&A on the interpretation of the fiscal 2024 medical fee schedule, part 1), office communication of 28 March 2024. Attachment 1 (medical fee schedule), Question 221, p. 医-58."", ""url"": ""https://www.mhlw.go.jp/content/12404000/001237675.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Pharmaceuticals and Medical Devices Agency (PMDA), Japan. 資料2, PMDA提出資料, 第22回 医療機器・体外診断薬の承認審査や安全対策等に関する定期意見交換会 (Document 2, submitted by PMDA to the 22nd regular exchange meeting on the review and safety of medical devices and in vitro diagnostics), 7 October 2024. Table AIを活用した医療機器(プログラム)の承認状況(R6.9末現在) (approval status of AI-based program devices as of the end of September 2024)."", ""url"": ""https://www.mhlw.go.jp/content/11124500/001329696.pdf""}]","{""kind"": ""Official documents""}" AIBD-0860,2024-02-08,event,day,reported,New Zealand's Privacy Commissioner finds that live facial recognition in 25 Foodstuffs North Island supermarkets reduced serious repeat offending,"New Zealand's Privacy Commissioner concluded that live facial recognition in 25 Foodstuffs North Island supermarkets reduced serious repeat offending, with an estimated 16% fewer serious incidents.",Live *Imagus* face matching cut serious repeat offending in 25 stores,an estimated 16% fewer serious incidents,,found,Found by a regulator,controlled comparison,very low,aggregate result,false,Imagus FRT System (live facial recognition),,"Imagus, an Australian-based developer according to the report",Australia,"Foodstuffs North Island, in 25 New World and PAK'nSAVE supermarkets",New Zealand,"North Island, New Zealand",New Zealand,Retail security,,Safety and rescue,Images and video,,16,High income,physical_health,Crime reduction or prevention,2,3,3,,3,5,,personal; organizational,1,0,"128 Foodstuffs North Island stores without facial recognition, over the same period",,Estimated 16% reduction in serious harmful behaviors and 21% reduction in shoplifting in the 25 trial stores,measured,persisted,2026-03-25,shown,https://www.privacy.org.nz/assets/DOCUMENTS/20250603-FRT-Inquiry-Report-A1082856.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Office of the Privacy Commissioner, New Zealand. Inquiry into Foodstuffs North Island trial use of facial recognition technology. Report dated May 2025, released 4 June 2025. Executive summary, para 8(b), p. 6. The Commissioner combined the evaluation that Foodstuffs commissioned with his office's own store visits, staff interviews and data requests."", ""url"": ""https://www.privacy.org.nz/assets/DOCUMENTS/20250603-FRT-Inquiry-Report-A1082856.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Office of the Privacy Commissioner, New Zealand. Inquiry into Foodstuffs North Island trial use of facial recognition technology, para 137, p. 39. Released 4 June 2025."", ""url"": ""https://www.privacy.org.nz/assets/DOCUMENTS/20250603-FRT-Inquiry-Report-A1082856.pdf""}]","{""kind"": ""Official documents""}" AIBD-0688,2024-01-01,event,year,mixed,Cour des comptes finds that reassessments from checks suggested by URSSAF data mining rose ninefold in Île-de-France from 2023 to 2024,The Cour des comptes found that reassessments from checks suggested by data mining rose ninefold at URSSAF Île-de-France. They went from €3.7 million in 2023 to €33.5 million in 2024.,URSSAF data mining lifts undeclared-work reassessments ninefold,rose ninefold at URSSAF Île-de-France,,found,Found by an auditor,before and after,very low,aggregate result,false,"ARTIC (Analyse des risques pour le travail d'investigation et le ciblage), the URSSAF data-mining tool",,Agence centrale des organismes de sécurité sociale (ACOSS),France,URSSAF network (figures from URSSAF Île-de-France),France,"Île-de-France, France",France,Social contributions enforcement (undeclared work),,Public sector,Prediction from data,Public servants,8; 16,High income,privacy_protection; operational_improvement,Fraud detection,2,1,4,,2,5,,organizational,1,0,Reassessments from checks suggested by data mining at URSSAF Île-de-France in 2023,"3,7 M€ in 2023","33,5 M€ in 2024",measured,unknown,,not shown,https://program-evaluation.ccomptes.fr/sites/default/files/2026-04/20260416-Lutte-fraude-prelevements-sociaux-et-travail-dissimule.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes. La politique de lutte contre la fraude aux prélèvements sociaux liée au travail dissimulé. Évaluation de politique publique, communication au comité d’évaluation et de contrôle des politiques publiques de l’Assemblée nationale, April 2026, pp. 76-77."", ""url"": ""https://program-evaluation.ccomptes.fr/sites/default/files/2026-04/20260416-Lutte-fraude-prelevements-sociaux-et-travail-dissimule.pdf""}]","{""kind"": ""Official documents""}" AIBD-0696,2024-01-01,event,year,reported,UNICEF's Evaluation Office finds that the I-HEAR-U listening platform informed UNICEF's 2024 flood response in Nepal,UNICEF's Evaluation Office found that the AI-powered I-HEAR-U platform informed UNICEF's 2024 flood response in Nepal. It informed HPV vaccine plans there. It cut staff work on reports.,*I-HEAR-U* informs UNICEF's response to Nepal's 2024 floods,,,found,Found by an evaluation office,no comparison,very low,program,false,I-HEAR-U (AI-powered social and community listening platform),,UNICEF Regional Office for South Asia with Research for Purpose,Nepal and Germany,Country offices of UNICEF (at least 5 in South Asia and 11 across UNICEF),,"South Asia, including Nepal and Pakistan",Nepal | Pakistan,Humanitarian response and health communication / social listening,,Humanitarian,Text and language,Public servants; Children,3; 13,Lower middle income,operational_improvement,"Faster data management, decision-making",2,3,,2,3,,,organizational; societal,1,0,,,,,unknown,,not shown,https://evaluationreports.unicef.org/GetDocument?documentID=19824&fileID=59675,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""UNICEF Evaluation Office. Evaluation of Innovation in UNICEF: South Asia Regional Thematic Case Study Report. New York, December 2025 (imprint November 2025). The case study was carried out between March and October 2025."", ""url"": ""https://evaluationreports.unicef.org/GetDocument?documentID=19824&fileID=59675""}]","{""kind"": ""Official documents""}" @@ -56,9 +56,9 @@ AIBD-0084,2022-01-01,event,year,reported,"DGFiP's “Foncier innovant” uses de AIBD-0705,2022-01-01,event,year,reported,Mexico's Superior Audit Office finds that an AI service built on Eviden's AIsaac reduced risk events at CFE in 2022,Mexico's Superior Audit Office checked reports from CFE's AIsaac-based security service. It found that AI correlation of events reduced risk events in 2022. It gave no count.,Mexican auditors find fewer cyber risk events at CFE with *AIsaac*,,,found,Found by an auditor,no comparison,very low,aggregate result,false,"Solución Avanzada de Detección de Amenazas y Respuesta por Inteligencia Artificial, Managed Detection and Response (MDR), delivered with Eviden's AIsaac tool",,"Eviden (Atos Group), maker of AIsaac. The service is supplied under contract 800961753 by Teliko Solutions, S.A.P.I. de C.V. and Grupo Vanguardia en Información y Conocimiento, S.A. de C.V.",France,"Comisión Federal de Electricidad (CFE), corporate office, with its subsidiaries and affiliates",Mexico,Mexico,Mexico,Cybersecurity,,Safety and rescue,Prediction from data,,9,Upper middle income,privacy_protection,Sensitive data management,2,1,4,,3,5,,organizational,1,0,,,,,unknown,,shown,https://www.asf.gob.mx/Trans/Informes/IR2022b/Documentos/Auditorias/2022_0040_a.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Auditoría Superior de la Federación. Informe Individual del Resultado de la Fiscalización Superior de la Cuenta Pública 2022. CFE Corporativo, Auditoría de Cumplimiento a Tecnologías de Información y Comunicaciones 2022-6-90UJB-20-0040-2023 (Auditoría núm. 40), Resultado 5. Opinion issued 13 October 2023."", ""url"": ""https://www.asf.gob.mx/Trans/Informes/IR2022b/Documentos/Auditorias/2022_0040_a.pdf""}]","{""kind"": ""Official documents""}" AIBD-0916,2022-01-01,event,year,reported,Rekenkamer Rotterdam finds that the share of parked cars that paid rose from about 50% to 90% in 2022 after the city's plate-reading scan cars came into use,The Rekenkamer Rotterdam found that the share of parked cars that paid rose from around 50% to 90% in 2022 after the city began checking with plate-reading scan cars.,*Parkeercontrole* scan cars lift Rotterdam's rate of paid parking to 90%,from around 50% to 90% in 2022,,found,Found by an auditor,before and after,very low,aggregate result,false,Parkeercontrole: scan cars (scanauto's) with an Arvoo image-recognition algorithm for license plates and the municipal ISA application,,"Arvoo (cameras and image-recognition software) with the Gemeente Rotterdam (ISA application, built by its ESO team at BCO)",Netherlands,"Gemeente Rotterdam, Stadsbeheer cluster, Parkeervoorzieningen department, Fiscale Handhaving team",Netherlands,"Rotterdam, Netherlands",Netherlands,Municipal parking enforcement,,Public sector,Images and video,Public servants,11,High income,operational_improvement,Enhanced productivity,2,1,4,,3,,,organizational,2,0,The payment rate before the scan cars came into use,rond de 50%,90% in 2022,measured,persisted,2026-02-20,shown,https://rekenkamer.rotterdam.nl/docs/onderzoeken/RO2205-kleur-bekennen-vervolgonderzoek-algoritmes-rekenkamer-rotterdam-casus-bijlagen.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Rekenkamer Rotterdam. Kleur bekennen: vervolgonderzoek algoritmes. Bijlage casusbevindingen, bevindingen casus parkeercontrole, §3-1 sturing en verantwoording and §3-2 model (PDF pp. 20-24). Published 13 May 2024."", ""url"": ""https://rekenkamer.rotterdam.nl/docs/onderzoeken/RO2205-kleur-bekennen-vervolgonderzoek-algoritmes-rekenkamer-rotterdam-casus-bijlagen.pdf""}]","{""kind"": ""Official documents""}" AIBD-0917,2022-01-01,event,year,reported,Netherlands Court of Audit finds that UWV's risk scan flagged files that held three times as much culpable unemployment as random files,The Netherlands Court of Audit found that UWV caseworkers find three times as much culpable unemployment in files flagged by the agency's risk scan as in a random sample.,*Risicoscan Verwijtbare Werkloosheid* triples the hit rate of UWV checks,17% of flagged files against 5.5% of random files,,found,Found by an auditor,controlled comparison,low,aggregate result,false,Risicoscan Verwijtbare Werkloosheid (risk scan for culpable unemployment),,UWV (Uitvoeringsinstituut Werknemersverzekeringen),Netherlands,"UWV, for unemployment benefit (WW) applications",Netherlands,Netherlands (national),Netherlands,Social security (unemployment benefit control),,Safety and rescue,Prediction from data,Public servants,16,High income,operational_improvement; privacy_protection,Enhanced productivity,2,1,4,,3,5,,organizational; personal,2,0,Unemployment benefit applications drawn at random and handled blind by the same caseworkers,"In 5,5% van de dossiers is verwijtbare werkloosheid gevonden door behandelaren",In 17% van de dossiers is verwijtbare werkloosheid gevonden door behandelaren,measured,persisted,2026-10-14,not shown,https://www.uwv.nl/nl/over-uwv/algoritmeregister-uwv/risicoscan-verwijtbare-werkloosheid,tussen de €29 en €44 miljoen per jaar,"Potential yearly benefits of working with the risk scan, by UWV's estimate",True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Algemene Rekenkamer. Resultaten verantwoordingsonderzoek 2024 ministerie van Sociale Zaken en Werkgelegenheid (XV), Rapport bij het Jaarverslag 2024, §5.1 Algoritme Risicoscan Verwijtbare Werkloosheid bij UWV, pp. 56-59, Figuur 8. Published 21 May 2025."", ""url"": ""https://www.rekenkamer.nl/site/binaries/site-content/collections/documents/2025/05/21/vo-2024-szw/XV+SZW+WR.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0918,2022-01-01,event,year,mixed,Brazil's Federal Court of Accounts finds that the FNDE's Malha Fina triage approved 27% of school-meal accounts but left the backlog unresolved,"Brazil's Federal Court of Accounts found that the FNDE's Malha Fina triage approved 11,931 school-meal accounts, 27% of those filed. Staff had concluded the analysis of only 3.6%.",*Malha Fina* clears 27% of school-meal accounts at Brazil's FNDE,"11,931 school-meal reports approved, 27% of the 44,386 filed",,found,Found by an auditor,,very low,aggregate result,false,"Malha Fina FNDE: an ABC value curve, rule-based audit trails and a supervised machine-learning Modelo Preditivo",,"Fundo Nacional de Desenvolvimento da Educação (FNDE), with information, data and technology exchanged with the Controladoria-Geral da União (CGU) under cooperation agreement FNDE-CGU 19/2020",Brazil,"Fundo Nacional de Desenvolvimento da Educação (FNDE), Diretoria Financeira (Difin)",Brazil,Brazil (national),Brazil,Public administration: education grant accountability review,,Public sector,Prediction from data,Public servants,16,Upper middle income,operational_improvement,"Faster data management, decision-making",2,1,1,,2,1,,organizational,1,0,School-meal accountability reports on which FNDE staff had concluded analysis by March 2023,"3,6% (1.369 contas) estavam com análise concluída",27% (11.931 contas) foram homologadas pelo Malha Fina,measured,unknown,,shown,https://contas.tcu.gov.br/sagas/SvlVisualizarRelVotoAc?codFiltro=SAGAS-SESSAO-ENCERRADA&seOcultaPagina=S&item0=852835,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Tribunal de Contas da União. Acórdão 743/2025 – TCU – Plenário (TC 020.760/2022-1), relator Ministro Walton Alencar Rodrigues, session of 2 April 2025. Operational audit of the FNDE's internal controls on the financial management of the Programa Nacional de Alimentação Escolar (PNAE), Relatório paragraphs 128-140 and Tabela 1."", ""url"": ""https://contas.tcu.gov.br/sagas/SvlVisualizarRelVotoAc?codFiltro=SAGAS-SESSAO-ENCERRADA&seOcultaPagina=S&item0=852835""}]","{""kind"": ""Official documents""}" +AIBD-0918,2022-01-01,event,year,mixed,Brazil's Federal Court of Accounts finds that the FNDE's Malha Fina triage approved 27% of school-meal accounts but left the backlog unresolved,"Brazil's Federal Court of Accounts found that the FNDE's Malha Fina triage approved 11,931 school-meal accounts, 27% of those filed. Staff had concluded the analysis of only 3.6%.",*Malha Fina* clears 27% of school-meal accounts at Brazil's FNDE,"11,931 school-meal reports approved, 27% of the 44,386 filed",,found,Found by an auditor,no comparison,very low,aggregate result,false,"Malha Fina FNDE: an ABC value curve, rule-based audit trails and a supervised machine-learning Modelo Preditivo",,"Fundo Nacional de Desenvolvimento da Educação (FNDE), with information, data and technology exchanged with the Controladoria-Geral da União (CGU) under cooperation agreement FNDE-CGU 19/2020",Brazil,"Fundo Nacional de Desenvolvimento da Educação (FNDE), Diretoria Financeira (Difin)",Brazil,Brazil (national),Brazil,Public administration: education grant accountability review,,Public sector,Prediction from data,Public servants,16,Upper middle income,operational_improvement,"Faster data management, decision-making",2,1,1,,2,1,,organizational,1,0,School-meal accountability reports on which FNDE staff had concluded analysis by March 2023,"3,6% (1.369 contas) estavam com análise concluída",27% (11.931 contas) foram homologadas pelo Malha Fina,measured,unknown,,shown,https://contas.tcu.gov.br/sagas/SvlVisualizarRelVotoAc?codFiltro=SAGAS-SESSAO-ENCERRADA&seOcultaPagina=S&item0=852835,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Tribunal de Contas da União. Acórdão 743/2025 – TCU – Plenário (TC 020.760/2022-1), relator Ministro Walton Alencar Rodrigues, session of 2 April 2025. Operational audit of the FNDE's internal controls on the financial management of the Programa Nacional de Alimentação Escolar (PNAE), Relatório paragraphs 128-140 and Tabela 1."", ""url"": ""https://contas.tcu.gov.br/sagas/SvlVisualizarRelVotoAc?codFiltro=SAGAS-SESSAO-ENCERRADA&seOcultaPagina=S&item0=852835""}]","{""kind"": ""Official documents""}" AIBD-1004,2022-01-01,event,year,reported,The Korea Communications Commission finds that the three mobile carriers' intelligent spam filters blocked 97.7% of spam texts in its tests in late 2022,"In tests by Korea's regulator, the three mobile carriers' free intelligent spam filters blocked 97.7% of illegal spam texts in the second half of 2022, up from 89.5% in late 2019.",Korean carriers' *intelligent spam blocking* stops 97.7% of spam tested,blocked 97.7% of illegal spam texts,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,"Intelligent spam blocking services (지능형 스팸차단 서비스) of SK Telecom, KT and LG U+",,"SK Telecom, KT and LG Uplus",South Korea,"SK Telecom, KT and LG Uplus (free service for subscribers)",South Korea,South Korea (nationwide mobile networks),South Korea,Telecommunications / spam and fraud prevention,,Finance,Text and language,Consumers,16,High income,privacy_protection,Fraud detection,2,,4,,3,,,personal,5,0,The same services' average blocking rate in the second half of 2019,89.5%,96.2% in the first half of 2020,measured,challenged,2024-04-05,shown,https://m.boannews.com/html/detail.html?idx=73575,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Regulatory document"", ""value"": ""Korea Communications Commission (방송통신위원회) and Korea Internet & Security Agency. 휴대전화 음성스팸 줄고, 문자스팸 늘어 [Voice spam on mobile phones falls, text spam rises], press release on the survey of spam in the second half of 2022 (「2022년 하반기 스팸 유통현황」), 31 March 2023."", ""url"": ""https://www.kmcc.go.kr/user.do?mode=view&page=A05030000&dc=K05030000&boardId=1113&boardSeq=54820""}, {""label"": ""Regulatory document"", ""value"": ""Korea Communications Commission (방송통신위원회) and Korea Internet & Security Agency. 방통위, 「2020년 상반기 스팸 유통현황」 조사 결과 발표 [The Commission announces its survey of spam in the first half of 2020], press release, 29 September 2020."", ""url"": ""https://www.kmcc.go.kr/user.do?mode=view&page=A05030000&dc=K05030000&boardId=1113&boardSeq=50090""}, {""label"": ""Regulatory document"", ""value"": ""Korea Communications Commission (방송통신위원회). 스팸 10건 중 9건은 막는다...문자스팸 차단율 89.5% [Nine in ten spam texts blocked, text spam blocking rate 89.5%], press release on the survey of spam in the second half of 2019 (「2019년 하반기 스팸 유통현황」), 25 March 2020."", ""url"": ""https://www.kmcc.go.kr/user.do?mode=view&page=A05030000&dc=K05030000&boardId=1113&boardSeq=48691""}, {""label"": ""Regulatory document"", ""value"": ""Korea Communications Commission (방송통신위원회) and Korea Internet & Security Agency. 방통위, 「2023년 하반기 스팸 유통현황」 발표 [The Commission announces its survey of spam in the second half of 2023], press release, 5 April 2024."", ""url"": ""https://www.kmcc.go.kr/user.do?mode=view&page=A05030000&dc=K05030000&boardId=1113&boardSeq=60499""}, {""label"": ""Regulatory document"", ""value"": ""Korea Communications Commission (방송통신위원회). 방통위(이상인 부위원장), 국민생활 불편을 야기하는 불법스팸 차단을 위해 이동통신사(KT) 현장점검 [The Commission's vice chairman inspects KT's blocking of illegal spam], press release, 20 February 2024."", ""url"": ""https://www.kmcc.go.kr/user.do?mode=view&page=A05030000&dc=K05030000&boardId=1113&boardSeq=59778""}, {""label"": ""Follow-up"", ""value"": ""The Commission reported that illegal spam texts received per person rose in the second half of 2023 and wrote that the carriers should strengthen the filtering of their intelligent (AI) spam blocking services."", ""url"": ""https://www.kmcc.go.kr/user.do?mode=view&page=A05030000&dc=K05030000&boardId=1113&boardSeq=60499""}]","{""kind"": ""Challenges""}" -AIBD-0691,2021-03-01,event,month,reported,GEF's evaluation office finds that AI monitoring of combustion kept dioxin at two Kunming incinerators well below the 0.1 ng TEQ/m³ target,"GEF's evaluation office found that AI monitoring kept dioxin at two Kunming incinerators far below target, as low as 0.0046 ng TEQ/m³. World Bank reviewers name no AI.",Kunming incinerators keep dioxin far below target with AI monitoring,as low as 0.0046 ng TEQ/m³,,found,Found by an evaluation office,,very low,benefit created,false,Combustion monitoring automated with artificial intelligence at the Xishan and Konggang incinerators (a SCADA system in the World Bank's review),,Not named in the evaluations,,"Operators of the Xishan and Konggang incinerators in Kunming, under the GEF-financed China Municipal Solid Waste Management Project implemented by the World Bank",China,"Kunming, Yunnan, China",China,Environment and climate,,Environment,Prediction from data,,12,Upper middle income,environmental_impact,Environmental impacts,2,2,4,,3,5,,societal,2,0,The 0.1 ng TEQ/m³ dioxin target,0.1 ng TEQ/m³ target,As low as 0.0046 ng TEQ/m³ at Xishan,measured,scaled,2025-11-19,not shown,https://www.unevaluation.org/get-download-file/18867,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Global Environment Facility Independent Evaluation Office. Evaluation of GEF Support for Technological Innovation. GEF/E/C.70/Inf.02, prepared by the Independent Evaluation Office for the 70th GEF Council Meeting, 19 November 2025. A field mission of the office visited the project."", ""url"": ""https://www.unevaluation.org/get-download-file/18867""}, {""label"": ""Evaluation report"", ""value"": ""World Bank Independent Evaluation Group. Implementation Completion Report (ICR) Review: CH GEF Municipal Solid Waste Management (P126832), ICRR0022775, 14 February 2022. In the review, the Independent Evaluation Group confirms the dioxin results, describes a SCADA system and does not mention artificial intelligence."", ""url"": ""https://documents.worldbank.org/curated/en/099350002142216484/pdf/China000CH0GEF0lid0Waste0Management.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0691,2021-03-01,event,month,reported,GEF's evaluation office finds that AI monitoring of combustion kept dioxin at two Kunming incinerators well below the 0.1 ng TEQ/m³ target,"GEF's evaluation office found that AI monitoring kept dioxin at two Kunming incinerators far below target, as low as 0.0046 ng TEQ/m³. World Bank reviewers name no AI.",Kunming incinerators keep dioxin far below target with AI monitoring,as low as 0.0046 ng TEQ/m³,,found,Found by an evaluation office,no comparison,very low,benefit created,false,Combustion monitoring automated with artificial intelligence at the Xishan and Konggang incinerators (a SCADA system in the World Bank's review),,Not named in the evaluations,,"Operators of the Xishan and Konggang incinerators in Kunming, under the GEF-financed China Municipal Solid Waste Management Project implemented by the World Bank",China,"Kunming, Yunnan, China",China,Environment and climate,,Environment,Prediction from data,,12,Upper middle income,environmental_impact,Environmental impacts,2,2,4,,3,5,,societal,2,0,The 0.1 ng TEQ/m³ dioxin target,0.1 ng TEQ/m³ target,As low as 0.0046 ng TEQ/m³ at Xishan,measured,scaled,2025-11-19,not shown,https://www.unevaluation.org/get-download-file/18867,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Global Environment Facility Independent Evaluation Office. Evaluation of GEF Support for Technological Innovation. GEF/E/C.70/Inf.02, prepared by the Independent Evaluation Office for the 70th GEF Council Meeting, 19 November 2025. A field mission of the office visited the project."", ""url"": ""https://www.unevaluation.org/get-download-file/18867""}, {""label"": ""Evaluation report"", ""value"": ""World Bank Independent Evaluation Group. Implementation Completion Report (ICR) Review: CH GEF Municipal Solid Waste Management (P126832), ICRR0022775, 14 February 2022. In the review, the Independent Evaluation Group confirms the dioxin results, describes a SCADA system and does not mention artificial intelligence."", ""url"": ""https://documents.worldbank.org/curated/en/099350002142216484/pdf/China000CH0GEF0lid0Waste0Management.pdf""}]","{""kind"": ""Official documents""}" AIBD-0840,2021-02-01,event,month,reported,The UK Covid-19 Inquiry found that QCovid's estimates of risk added 1.7 million people to England's Shielded Patient List in February 2021,"The UK Covid-19 Inquiry reports that QCovid data led to 1.7 million more people on England's Shielded Patient List in February 2021. Of those with ethnicity recorded, 36% were non-white.",*QCovid* adds 1.7 million people to England's shielding list,1.7 million more people on England's Shielded Patient List,,found,Found by an inquiry,no comparison,very low,aggregate result,false,QCovid risk stratification tool,,"Researchers led from the University of Oxford, using the QResearch database of general practice records",United Kingdom,"NHS England with NHS Digital, which kept the Shielded Patient List in England",United Kingdom,"England, United Kingdom",United Kingdom,Public health / pandemic risk stratification,,Health,Prediction from data,Patients,3,High income,physical_health,Medical access,2,5,2,4,3,1,0.260,personal; societal,2,0,The adult population,17% non-white ethnicity recorded in the adult population,36% were non-white,measured,persisted,,shown,https://doi.org/10.1136/bmj.m3731,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Inquiry report"", ""value"": ""UK Covid-19 Inquiry (Chair: The Rt Hon the Baroness Hallett DBE). Module 3: The impact of the Covid-19 pandemic on the healthcare systems of the United Kingdom (HC 1730), ordered to be printed 19 March 2026, paras 3.35-3.38. Accessible PDF with the May 2026 correction slip."", ""url"": ""https://covid19.public-inquiry.uk/wp-content/uploads/2026/03/14162519/14.5.26-E03361629_Covid-Inquiry-M3-Inquiry_accessible-w-Corrslip-2.pdf""}]","{""kind"": ""Official documents""}" AIBD-0094,2021-01-01,event,month,reported,"Nigeria's Rapid Response Register used machine-learning maps of poverty to select 1,163 urban wards for cash transfers in response to COVID-19","Nigeria used machine-learning poverty maps from satellite imagery to select 1,163 of its poorest urban wards for a COVID-19 cash transfer that counted 49,000 beneficiary households by 15 October 2021.",Nigeria uses machine-learning poverty maps to direct COVID-19 cash aid,"49,000 beneficiary households as of 15 October 2021","Nigeria had no urban social registry. Its National Social Safety Nets Coordinating Office therefore built the Rapid Response Register for a government-funded cash transfer during the COVID-19 pandemic. The government used maps of poverty by ward, made by applying machine-learning algorithms to satellite imagery. With them it selected 1,163 of the poorest urban wards, with a population of over 20 million.[1] Their residents then registered by mobile phone and were verified in person. The developers of the maps report that, in targeting the poor, their machine-learning approach raises precision at the ward level to 0.92. A survey-based benchmark reaches 0.79.[2] No source shows how many people the maps helped. A World Bank status report of 29 January 2025 records 49,000 beneficiary households in the cash transfer as of 15 October 2021. The World Bank does not say how many of them had been paid. It also counts over 5 million households supported by the program of direct benefit transfers. It gives no breakdown of the share that the register targeted.[3] Human Rights Watch researchers argued in July 2021 that the targeting methods had in all likelihood excluded, or would exclude, large numbers of people in need.[4]",found,Found by an evaluation office,tested against a standard,low,benefit created,false,"Ward-level poverty maps from supervised machine learning (deep learning) on satellite imagery, used for geographic targeting in the Rapid Response Register",,"Joshua Blumenstock (UC Berkeley) and Isabella Smythe (Columbia University), using wealth estimates by Chi et al. (2021), with World Bank economists Jonathan Lain and Tara Vishwanath, for the Government of Nigeria",,"National Social Safety Nets Coordinating Office (NASSCO), Federal Government of Nigeria",Nigeria,Urban wards across Nigeria,Nigeria,Social protection and cash transfers,Government social assistance,Humanitarian,Images and video,The public,1,Lower middle income,opportunity_access,Insurance and benefit access,2,4,3,4,4,3,0.455,personal; organizational; societal,6,1,survey-based poverty map benchmark (DHS/NLSS data),0.79 precision at the ward level (survey benchmark),0.92 precision at the ward level (ML-based maps),measured,scaled,2025-01-29,shown,https://blogs.worldbank.org/en/opendata/using-big-data-and-machine-learning-locate-poor-nigeria,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""World Bank Independent Evaluation Group. Implementation Completion Report Review: National Social Safety Nets Project (P151488), ICRR0023725, 26 January 2024. The evaluation group reports to the World Bank's Board and validates the results that project teams report."", ""url"": ""https://documents.worldbank.org/curated/en/099012624183017071/pdf/P1514881e50fcd0261bd211ca45ab33dadc.pdf""}]","{""kind"": ""Official documents""}" AIBD-0683,2021-01-01,event,year,reported,"Cour des comptes finds that France Travail's LEGO model catches illegal job offers, cutting checking work by the equivalent of 98 posts",The Cour des comptes found that France Travail's LEGO model catches illegal job offers before publication. It cut advisers' checking work by the equivalent of 98 full-time posts.,*LEGO* catches illegal job offers before France Travail publishes them,93% of offers that LEGO flags are flagged correctly,,found,Found by an auditor,tested against a standard,low,aggregate result,false,"LEGO (Légalité des offres), natural language processing with deep neural networks",,"Pôle emploi, now France Travail",France,France Travail (business advisers in all agencies),France,France (national),France,Public employment services,,Public sector,Text and language,Public servants,8,High income,operational_improvement,Enhanced productivity,2,1,4,,3,5,,organizational; societal,1,0,"TextKernel, the keyword software used until autumn 2021, run in parallel during 2021",« seulement » 65 % avec l’ancien logiciel,93 % des offres d’emploi signalées par LEGO le sont à raison,measured,unknown,,shown,https://www.ccomptes.fr/sites/default/files/2026-01/20260108-S2025-1558-France-Travail-et-intelligence-artificielle_0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes. France Travail et l’intelligence artificielle. Observations définitives S2025-1558, adopted 15 October 2025 and published 8 January 2026, p. 9 and annex 5, pp. 111-112."", ""url"": ""https://www.ccomptes.fr/sites/default/files/2026-01/20260108-S2025-1558-France-Travail-et-intelligence-artificielle_0.pdf""}]","{""kind"": ""Official documents""}" @@ -66,7 +66,7 @@ AIBD-0735,2020-12-01,source publication,day,reported,"NICE finds that Zio XT, a AIBD-0093,2020-11-02,event,day,reported,"Togo's Novissi used machine learning on satellite imagery to send emergency cash to rural areas, where phone data identified recipients","Machine-learning models used satellite imagery to rank rural areas for Togo's Novissi cash. In those areas, phone metadata set who got it. From November 2020 to August 2021, 138,531 people were paid.",Togo uses machine learning to send emergency cash to its rural poor,reduced exclusion errors by 4–21% relative to geographic alternatives,,found,Found by an evaluation office,tested against a standard,low,aggregate result,false,Novissi rural targeting pipeline: satellite-imagery poverty maps for canton ranking plus supervised machine learning on anonymized mobile-phone call detail records for individual eligibility,gradient boosting regressor with Microsoft's LightGBM,"Innovations for Poverty Action, CEGA and UC Berkeley (Aiken and Blumenstock, UC Berkeley; Bellue, University of Mannheim; Karlan and Udry, Northwestern University) for the Government of Togo",United States,"Government of Togo, Ministry of Digital Economy and Transformation, with GiveDirectly",Togo,"Rural cantons, Togo",Togo,Social protection and humanitarian cash transfers,Government social assistance,Humanitarian,Prediction from data; Images and video,The public,1,Lower middle income,opportunity_access,Insurance and benefit access,2,4,3,4,4,3,0.455,personal; organizational; societal,9,1,the geographic targeting options considered by the Government of Togo,"errors of exclusion above 53% for the feasible alternatives, by the 4–21% that the authors report","errors of exclusion of 53% for the phone-based approach, a reduction of 4–21%",measured,scaled,2022-01-24,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC8967719/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Independent Evaluation Group (IEG), World Bank Group (2023). Togo - Togo Emergency Covid-19 DPO 2021 (P174376): Implementation Completion Report (ICR) Review, Section 12, Lessons, pages 19-20 (lesson 1 on digitalization); results indicator RI3 in Section 4."", ""url"": ""http://documents.worldbank.org/curated/en/099120523130534785/pdf/P1743760994a770790a2420969df01036f2.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Independent Evaluation Group (IEG), World Bank Group (2025). How the World Bank Supports Adaptive Social Protection in Crisis Response - An Independent Evaluation, Chapter 2, box on expanding safety nets in crises, bullet 'Using alternative data sources' (source line: IEG structured literature review)."", ""url"": ""https://documents.worldbank.org/curated/en/099420007082597981/pdf/SECBOS-edce0da9-4c32-4e0a-8a68-8cdf5bea0fee.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Independent Evaluation Group (IEG), World Bank Group (2022). The World Bank's Early Support to Addressing Coronavirus (COVID-19) Health and Social Response - An Early-Stage Evaluation, Chapter 4, box of examples of partnerships in the social response (Togo paragraph)."", ""url"": ""http://documents.worldbank.org/curated/en/099707411152235594/pdf/SECBOS0a7075fa02d096800f25989934099.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Independent Evaluation Group (IEG), World Bank Group (2024). Togo - TG-Safety Nets and Basic Services Project (P157038): Implementation Completion Report (ICR) Review, Section 4, Objective 1 rationale, page 8."", ""url"": ""https://documents.worldbank.org/curated/en/099102424140039295/pdf/P157038-8781f09e-ea59-4a0d-bbd1-e3d699e00992.pdf""}]","{""kind"": ""Official documents""}" AIBD-0651,2020-09-18,source publication,day,reported,CMS approves add-on payments for Viz.ai ContaCT after determining that it is a substantial clinical improvement because it shortens stroke-specialist notification,"CMS found that Viz.ai ContaCT shortens time to specialist notification for large vessel occlusion stroke. In routine use, the median time to notification was 4.31 minutes.",Medicare credits *ContaCT* with faster stroke alerts in a payment review,the median time to notification was 4.31 minutes,,found,Found by a regulator,tested against a standard,low,aggregate result,false,ContaCT (Viz LVO algorithm with Viz Hub and Viz View),,"Viz.ai, Inc.",United States,"US hospitals using ContaCT, including SkyRidge Medical Center, Wellstar, Methodist Le Bonheur Healthcare and Mount Sinai Health System",United States,United States,United States,Healthcare,,Health,Images and video,Health workers; Patients,3,High income,physical_health,Medical diagnosis,2,3,3,,5,,,organizational; personal,2,3,"Usual-care notification time in the retrospective test on past scans that FDA reviewed, as cited by CMS",58.72 minutes mean,7.32 minutes mean,measured,persisted,2021-08-04,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN170073.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Centers for Medicare & Medicaid Services. Medicare Program; Hospital Inpatient Prospective Payment Systems for Acute Care Hospitals and the Long-Term Care Hospital Prospective Payment System; final rule. Federal Register 85 FR 58432, 18 September 2020 (document 2020-19637)."", ""url"": ""https://www.federalregister.gov/documents/2020/09/18/2020-19637/medicare-program-hospital-inpatient-prospective-payment-systems-for-acute-care-hospitals-and-the-long-term-care-hospital""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration. De Novo classification request for ContaCT (DEN170073): decision summary. 13 February 2018."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN170073.pdf""}]","{""kind"": ""Official documents""}" AIBD-0999,2020-07-15,event,day,reported,The CRTC approves Bell Canada's AI-assisted call blocking after it blocked over one billion fraudulent calls with no confirmed false positives,"Bell Canada reported to the CRTC that its AI-assisted system blocked 1,120,372,443 calls in 15 months. The CRTC judged the risk of wrongful blocking minimal. It made the blocking permanent.",Bell uses AI to block over a billion fraudulent calls in Canada,"1,120,372,443 calls",,found,Found by a regulator,no comparison,very low,aggregate result,false,Bell Canada's network-level call-blocking system for fraudulent and scam voice calls (AI anomaly flagging with review),,Bell Canada,Canada,Bell Canada and its affiliates (Bell Canada et al.),Canada,Canada,Canada,Telecommunications / fraud prevention,,Finance,Prediction from data,The public,16,High income,privacy_protection,Fraud detection,2,,4,3,3,5,,societal,3,0,Complaints about fraudulent calls received by Bell and the CRTC before the trial began,,,measured,persisted,2022-09-29,shown,https://business.bell.ca/support/medium-large/mobility/fraudulent-abusive-calls,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Canadian Radio-television and Telecommunications Commission. Compliance and Enforcement and Telecom Decision CRTC 2021-403: Bell Canada - Application to allow Bell Canada and its affiliates to block certain fraudulent and scam voice calls on a permanent basis. Ottawa, 9 December 2021. Archived copy of https://www.crtc.gc.ca/eng/archive/2021/2021-403.htm."", ""url"": ""https://web.archive.org/web/20260520150034id_/https://crtc.gc.ca/eng/archive/2021/2021-403.htm""}, {""label"": ""Regulatory document"", ""value"": ""Canadian Radio-television and Telecommunications Commission. Compliance and Enforcement and Telecom Decision CRTC 2021-403. Ottawa, 9 December 2021. Archived copy of https://www.crtc.gc.ca/eng/archive/2021/2021-403.htm."", ""url"": ""https://web.archive.org/web/20260520150034id_/https://crtc.gc.ca/eng/archive/2021/2021-403.htm""}, {""label"": ""Regulatory document"", ""value"": ""Canadian Radio-television and Telecommunications Commission. Compliance and Enforcement and Telecom Decision CRTC 2021-403. Ottawa, 9 December 2021. Archived copy of https://www.crtc.gc.ca/eng/archive/2021/2021-403.htm."", ""url"": ""https://web.archive.org/web/20260520150034id_/https://crtc.gc.ca/eng/archive/2021/2021-403.htm""}]","{""kind"": ""Official documents""}" -AIBD-0297,2020-01-01,event,year,reported,"Nelli, an audiovisual AI monitor for seizures, records 247 nocturnal events missed by carers among 15 people with intellectual disability in Cornwall. In a Danish validation it detected all 11 tonic-clonic seizures","In a Cornwall NHS evaluation of 15 people who could not tolerate EEG, Nelli flagged 707 possible seizure events, 247 unheard or unrecognized by carers. Danish validators reported 93.7% sensitivity.",Nelli flags 247 nocturnal seizure events missed by carers,"Nelli flagged 707 possible seizure events, 247 unheard or unrecognized by carers",,found,Found by an assessment body,,very low,aggregate result,false,Nelli audio-video seizure monitoring system (computer vision and AI-based event detection),,"Neuro Event Labs, Tampere, Finland",Finland,"Cornwall Partnership NHS Foundation Trust (community evaluation); Danish Epilepsy Center, Dianalund, and Aarhus University Hospital (validation)",United Kingdom,"Cornwall, United Kingdom; Dianalund and Aarhus, Denmark",United Kingdom | Denmark,Healthcare (epilepsy; nocturnal seizure monitoring),Residential and institutional seizure monitoring for people with intellectual disability,Accessibility,Images and video; Speech and sound,Patients; People with disabilities,3,High income,physical_health,Medical diagnosis,2,2,2,4,5,5,0.500,personal; organizational,8,3,carers' hearing or recognition of night-time events in the same people,247 were not heard or recognised by carers,Nelli recorded 707 possible seizure events across the study cohort,measured,persisted,2025-10-09,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf25/K251506.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, 510(k) premarket notification K251506, Nelli (Version 7.11), decision of 21 November 2025, with the 510(k) summary written by Neuro Event Labs. The summary covers the sponsor's clinical study at three U.S. sites. FDA found the device substantially equivalent to the Empatica Embrace."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf25/K251506.pdf""}, {""label"": ""Assessment report"", ""value"": ""Health Technology Wales (HTW) (2025). HTW Guidance 064: At-home video monitoring for the diagnosis and management of epilepsy, p. 1-2, Evidence summary."", ""url"": ""https://healthtechnology.wales/wp-content/uploads/GUI064-AVME-WEB-ENG.pdf""}, {""label"": ""Assessment report"", ""value"": ""Health Technology Wales, Evidence Appraisal Report EAR064: At-home video monitoring for the diagnosis and management of epilepsy, August 2025, section 5.5.5 (diagnostic accuracy: false positive rate)."", ""url"": ""https://healthtechnology.wales/wp-content/uploads/EAR064-AVME-WEB.pdf""}, {""label"": ""Assessment report"", ""value"": ""Health Technology Wales, Guidance 064 (GUI064): At-home video monitoring for the diagnosis and management of epilepsy, August 2025, headline guidance."", ""url"": ""https://healthtechnology.wales/wp-content/uploads/GUI064-AVME-WEB-ENG.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0297,2020-01-01,event,year,reported,"Nelli, an audiovisual AI monitor for seizures, records 247 nocturnal events missed by carers among 15 people with intellectual disability in Cornwall. In a Danish validation it detected all 11 tonic-clonic seizures","In a Cornwall NHS evaluation of 15 people who could not tolerate EEG, Nelli flagged 707 possible seizure events, 247 unheard or unrecognized by carers. Danish validators reported 93.7% sensitivity.",Nelli flags 247 nocturnal seizure events missed by carers,"Nelli flagged 707 possible seizure events, 247 unheard or unrecognized by carers",,found,Found by an assessment body,controlled comparison,low,aggregate result,false,Nelli audio-video seizure monitoring system (computer vision and AI-based event detection),,"Neuro Event Labs, Tampere, Finland",Finland,"Cornwall Partnership NHS Foundation Trust (community evaluation); Danish Epilepsy Center, Dianalund, and Aarhus University Hospital (validation)",United Kingdom,"Cornwall, United Kingdom; Dianalund and Aarhus, Denmark",United Kingdom | Denmark,Healthcare (epilepsy; nocturnal seizure monitoring),Residential and institutional seizure monitoring for people with intellectual disability,Accessibility,Images and video; Speech and sound,Patients; People with disabilities,3,High income,physical_health,Medical diagnosis,2,2,2,4,5,5,0.500,personal; organizational,8,3,carers' hearing or recognition of night-time events in the same people,247 were not heard or recognised by carers,Nelli recorded 707 possible seizure events across the study cohort,measured,persisted,2025-10-09,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf25/K251506.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, 510(k) premarket notification K251506, Nelli (Version 7.11), decision of 21 November 2025, with the 510(k) summary written by Neuro Event Labs. The summary covers the sponsor's clinical study at three U.S. sites. FDA found the device substantially equivalent to the Empatica Embrace."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf25/K251506.pdf""}, {""label"": ""Assessment report"", ""value"": ""Health Technology Wales (HTW) (2025). HTW Guidance 064: At-home video monitoring for the diagnosis and management of epilepsy, p. 1-2, Evidence summary."", ""url"": ""https://healthtechnology.wales/wp-content/uploads/GUI064-AVME-WEB-ENG.pdf""}, {""label"": ""Assessment report"", ""value"": ""Health Technology Wales, Evidence Appraisal Report EAR064: At-home video monitoring for the diagnosis and management of epilepsy, August 2025, section 5.5.5 (diagnostic accuracy: false positive rate)."", ""url"": ""https://healthtechnology.wales/wp-content/uploads/EAR064-AVME-WEB.pdf""}, {""label"": ""Assessment report"", ""value"": ""Health Technology Wales, Guidance 064 (GUI064): At-home video monitoring for the diagnosis and management of epilepsy, August 2025, headline guidance."", ""url"": ""https://healthtechnology.wales/wp-content/uploads/GUI064-AVME-WEB-ENG.pdf""}]","{""kind"": ""Official documents""}" AIBD-0634,2019-11-27,event,day,reported,"PROSPECT randomized trial in Blantyre, Malawi: HIV testing combined with CAD4TB-read chest X-rays speeds TB treatment for adults with cough","In the PROSPECT trial in Blantyre, TB patients offered HIV testing with chest X-rays read by CAD4TB started treatment a median of 1 day after the visit. Under standard care, they waited 11 days.",Computer-read chest X-rays speed TB treatment at a Malawi clinic,Median time to treatment was 1 day in the X-ray arm and 11 days under standard care,,found,Found by an assessment body,randomized trial,low,aggregate result,false,CAD4TB computer-aided detection software reading digital chest X-rays,CAD4TB v5,Delft Imaging,Netherlands,PROSPECT trial team of the Malawi-Liverpool-Wellcome Trust Clinical Research Programme at Bangwe Health Centre,Malawi,"Bangwe Health Centre, Blantyre, Malawi",Malawi,Healthcare,,Health,Images and video,Patients,3,Low income,physical_health,Medical diagnosis,2,2,1,4,3,1,0.180,personal,2,1,Standard care for adults with cough at the same clinic,0 of 5 TB patients treated on the day of the visit; median 11 days to treatment,6 of 15 TB patients treated on the day of the visit; median 1 day to treatment,measured,unknown,,shown,https://iris.who.int/server/api/core/bitstreams/0d7e8935-879e-4ab6-8c07-e41f7be1465b/content,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Malaysian Health Technology Assessment Section (MaHTAS), Medical Development Division, Ministry of Health Malaysia (2024). Computer-aided detection or artificial intelligence for chest X-ray in early detection of tuberculosis: Health Technology Assessment, executive summary, Results, Part A: Diagnostic accuracy and effectiveness (p. 3); Conclusion and Recommendation (p. 5)."", ""url"": ""https://web.archive.org/web/20260118050712id_/https://www.moh.gov.my/moh/resources/Penerbitan/Penilaian%20Teknologi%20Kesihatan%20(HTA)/ES_COMPUTER_AIDED-DETECTION_OR_ARTIFICIAL_INTELLIGENCE_FOR_CHEST_X-RAY_IN_EARLY_DETECTION_OF_TUBERCULOSIS.pdf""}]","{""kind"": ""Official documents""}" AIBD-0623,2019-11-01,event,month,reported,FDA De Novo review of GI Genius: AI-aided colonoscopy detects more adenomas in a randomized trial,"In the randomized Italian trial that FDA reviewed, colonoscopy with GI Genius found adenomas in 55.1% of low-risk patients, against 42.0% with standard colonoscopy.",*GI Genius* helped endoscopists find more adenomas in a randomized trial,an adenoma detection rate of 55.1% with GI Genius against 42.0% with standard colonoscopy,,found,Found by a regulator,randomized trial,high,aggregate result,false,GI Genius,,"Cosmo Artificial Intelligence - AI, Ltd.",Ireland,"Humanitas Research Hospital, Nuovo Regina Margherita Hospital and Valduce Hospital (AID trial)",Italy,"Milan, Rome and Como, Italy",Italy,Healthcare,,Health,Images and video,Health workers; Patients,3,High income,physical_health,Medical diagnosis,2,3,1,,3,1,,personal; organizational,6,0,Standard white-light colonoscopy in the same trial,42.0% adenoma detection rate,55.1% adenoma detection rate,measured,persisted,2026-05-29,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200055.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for GI Genius (DEN200055): decision summary. Granted 2021-04-09."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200055.pdf""}, {""label"": ""Assessment report"", ""value"": ""Agència de Qualitat i Avaluació Sanitàries de Catalunya (AQuAS), for the Spanish Network of HTA Agencies (RedETS) and the Ministry of Health (2023). Inteligencia artificial para la detección y caracterización de lesiones precancerosas colorrectales en la colonoscopia (Ficha de evaluación de tecnologías sanitarias nuevas y emergentes), Resumen, Resultados (PDF p. 14; repeated p. 52)."", ""url"": ""https://redets.sanidad.gob.es/productos/buscarProductos.do?metodo=descargarFichero&idProducto=1177""}, {""label"": ""Assessment report"", ""value"": ""Danish Health Technology Council (Behandlingsrådet), expert committee on AI in colonoscopy (2023). Recommendation from the Danish Health Technology Council concerning use of artificial intelligence as clinical decision-support in colonoscopy for the diagnosis of neoplastic disease, Clinical effectiveness (p. 3 of 7); the analysis report (analyserapport_1.pdf) names GI Genius and ENDO-AID as the systems in the evidence."", ""url"": ""https://behandlingsraadet-dk.b-cdn.net/media/35dpf1vw/recommendation.pdf""}, {""label"": ""Assessment report"", ""value"": ""Agència de Qualitat i Avaluació Sanitàries de Catalunya (AQuAS), for the Spanish HTA network (RedETS) and Ministry of Health (2023). Inteligencia artificial para la detección y caracterización de lesiones precancerosas colorrectales en la colonoscopia (ficha de evaluación de tecnologías nuevas y emergentes), Resumen, Resultados (p. 13-14)."", ""url"": ""https://scientiasalut.gencat.cat/bitstream/handle/11351/10545/inteligencia_artificial_deteccion_caracterizacion_lesiones_precancerosas_colorrectales_colonoscopia_2023.pdf?sequence=1&isAllowed=y""}]","{""kind"": ""Official documents""}" AIBD-1029,2019-06-11,source publication,day,reported,Netherlands Court of Audit finds that the Belastingdienst's risk model for VAT refunds doubled the hit rate of the selection rules that it replaced,The Netherlands Court of Audit found that the Belastingdienst's risk model for VAT refunds raised the hit rate of checks from 17% to 34%.,*Risicomodel van teruggave btw* doubles the hit rate of VAT checks,from 17% to 34%,,found,Found by an auditor,before and after,very low,aggregate result,false,Risicomodel van teruggave btw (the Belastingdienst's risk model for VAT refund returns),,Belastingdienst (Netherlands Tax and Customs Administration),Netherlands,"Belastingdienst, for the selection of VAT refund returns for checks by tax inspectors",Netherlands,Netherlands (national),Netherlands,Tax compliance,,Public sector,Prediction from data,Public servants,16,High income,operational_improvement; privacy_protection,Value above replacement,2,1,4,,3,5,,organizational; societal,2,0,"The old selection rules for VAT refund returns, which the model replaced",De trefkans van het oude model voor btw-teruggave was 17%,met het nieuwe model wordt een trefkans van 34% gehaald,measured,litigated,2026-04-22,shown,https://www.rekenkamer.nl/site/binaries/site-content/collections/documents/2019/06/11/datagedreven-selectie-van-aangiften-door-de-belastingdienst/Dategedreven+selectie+aangiften+door+Belastingdienst+WR.pdf,€ 60 miljoen per jaar,"Extra revenue from the new VAT refund model, by the Belastingdienst's estimate",True,[],Querist,The use of the system was taken to court.,"[{""label"": ""Audit report"", ""value"": ""Algemene Rekenkamer (Netherlands Court of Audit). Datagedreven selectie van aangiften door de Belastingdienst, §2.1.3 Trefkans, p. 10, and Bijlage 1 'Van data naar risicomodel', pp. 28-29. Sent to the House of Representatives on 11 June 2019."", ""url"": ""https://www.rekenkamer.nl/site/binaries/site-content/collections/documents/2019/06/11/datagedreven-selectie-van-aangiften-door-de-belastingdienst/Dategedreven+selectie+aangiften+door+Belastingdienst+WR.pdf""}, {""label"": ""Follow-up"", ""value"": ""The Belastingdienst switched off its OB Negatief model for VAT refund returns in January 2021 over privacy doubts and later restored it. In April 2026 the District Court of The Hague ruled its use in a VAT dispute lawful."", ""url"": ""https://www.accountancyvanmorgen.nl/2026/04/22/risicoselectie-fiscus-via-algoritme-rechtmatig-naheffingen-en-boetes-administratiekantoor/""}]","{""kind"": ""Legal proceedings""}" @@ -78,7 +78,7 @@ AIBD-0856,2019-01-01,event,month,reported,Australia's Administrative Review Trib AIBD-0785,2018-08-23,source publication,day,reported,FDA finds that Zephyr's MATRx plus predicted which patients with sleep apnea would respond to an oral appliance with 94.2% sensitivity,"In a home study that FDA reviewed, MATRx plus identified patients with mild to moderate sleep apnea who would respond to an oral appliance with 94.2% sensitivity. Its PPV was 96.1%.",*MATRx plus* identified apnea patients who did well with a mouthpiece,94.2% sensitivity,,found,Found by a regulator,no comparison,very low,aggregate result,false,MATRx plus,,Zephyr Sleep Technologies,Canada,Zephyr Sleep Technologies (home study in adults with mild to moderate obstructive sleep apnea),Canada,,,Healthcare,,Health,Prediction from data,Patients,3,,physical_health,,2,2,1,,3,1,,organizational,3,0,,,,,withdrawn,2021-07-21,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN170090.pdf,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo Classification Request for MATRx plus (DEN170090): decision summary. Granted 2018-08-23."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN170090.pdf""}, {""label"": ""Follow-up"", ""value"": ""Zephyr Sleep Technologies ceased operations on 25 June 2021, closing manufacturing and support for MATRx plus after failing to raise capital during the pandemic, Sleep Review reported on 21 July 2021."", ""url"": ""https://sleepreviewmag.com/sleep-treatments/therapy-devices/oral-appliances/zephyr-sleep-technologies-ceases-operations/""}]","{""kind"": ""Withdrawals""}" AIBD-0223,2018-08-22,event,day,reported,CBP facial comparison at Dulles flags a French-passport impostor with a Congolese ID in his shoe on day three,On 22 August 2018 CBP's new facial comparison system at Dulles flagged a mismatch for a man presenting a French passport. Officers found a Congolese ID in his shoe. He was returned to Brazil.,CBP facial comparison at Dulles flags a French-passport impostor,the first impostor detected with the technology,,found,Found by an auditor,no comparison,very low,harm prevented,false,"CBP facial comparison biometric system (biometric entry/exit, Traveler Verification Service)",,Not named in the sources,,"U.S. Customs and Border Protection, Office of Field Operations, Washington Dulles International Airport",United States,"Sterling, United States",United States,Border security,Government,Public sector,Images and video,Public servants,16,High income,privacy_protection; physical_health,Fraud detection,2,1,1,,5,3,,societal; organizational,6,3,an officer's visual check of the traveler against the passport photo,officials said that the passport may have passed at face value with human inspection,the facial comparison reported a mismatch and the traveler was sent to secondary inspection,assumed,scaled,2022-07-05,not shown,https://www.gao.gov/assets/gao-20-568.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""U.S. Government Accountability Office, Facial Recognition: CBP and TSA are Taking Steps to Implement Programs, but CBP Should Address Privacy and System Performance Issues, GAO-20-568, September 2020. GAO is an independent audit office. On impostors it reports only what CBP officials told it. Its own findings cover privacy notices and photo capture in the exit program."", ""url"": ""https://www.gao.gov/assets/gao-20-568.pdf""}, {""label"": ""Audit report"", ""value"": ""U.S. Department of Homeland Security, Office of Inspector General. CBP Complied with Facial Recognition Policies to Identify International Travelers at Airports. Report OIG-22-48, 5 July 2022. Results of Audit and Appendix B, 'Impostors Identified by CBP at Airports'. The inspector general analysed CBP's own records, and the audit period begins after this case."", ""url"": ""https://www.oversight.gov/sites/default/files/documents/reports/2022-07/OIG-22-48-July22.pdf""}]","{""kind"": ""Official documents""}" AIBD-0920,2018-01-01,event,year,reported,"Cour des comptes finds that when data mining picked very small firms for URSSAF audits, the audits reassessed contributions at twice the rate of all such audits",The Cour des comptes found that URSSAF audits of very small firms picked by data mining reassessed 12.6% of the contributions audited in 2018. The rate across all such audits was 6.1%.,*Datamining* lists double the yield of URSSAF audits of small firms,reassessed 12.6% of the contributions audited in 2018,,found,Found by an auditor,controlled comparison,very low,aggregate result,false,Acoss data-mining risk scores (outils de datamining) that list small and medium firms (since 2015) and very small firms (since 2016) for URSSAF accounting audits,,Agence centrale des organismes de sécurité sociale (Acoss),France,URSSAF network,France,France (national),France,Social contributions enforcement (employer audits),,Public sector,Prediction from data,,16,High income,operational_improvement; privacy_protection,Enhanced productivity,2,1,4,,3,,,organizational,1,0,"All URSSAF accounting audits of very small firms in 2018, including those picked by data mining","6,1 %","12,6 %",measured,scaled,2024-05-28,not shown,https://www.ccomptes.fr/sites/default/files/2023-10/20191202-rapport-fraude-aux-prelevements-obligatoires.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes. La fraude aux prélèvements obligatoires: évaluer, prévenir, réprimer. Communication au Premier ministre, November 2019, published 2 December 2019, pp. 128-129, Tableau n° 14."", ""url"": ""https://www.ccomptes.fr/sites/default/files/2023-10/20191202-rapport-fraude-aux-prelevements-obligatoires.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0675,2017-08-01,event,year,mixed,DHS Inspector General validates that the Traveler Verification Service matched 98% of gate photos in CBP's 2017 exit pilot at nine airports,"The DHS Inspector General validated a 98% average match rate for CBP's face matching at departure gates from August to December 2017, but only 85% of passengers were confirmed.",*Traveler Verification Service* matched 98% of photographed passengers,a 98% average match rate,,found,Found by an auditor,,very low,aggregate result,false,"Traveler Verification Service (TVS), CBP's facial recognition matching service",,U.S. Customs and Border Protection (the Inspector General does not name the algorithm's supplier),United States,"U.S. Customs and Border Protection, Biometric Entry-Exit Program, with airlines at nine airports (Sprint 8 pilot)",United States,"Atlanta, Boston, Chicago, Houston (Intercontinental and Hobby), Las Vegas, Miami, New York JFK and Washington Dulles airports, United States",United States,Border control (exit records),,Safety and rescue,Images and video,,16,High income,operational_improvement,"Faster data management, decision-making",2,1,3,2,2,5,0.280,organizational,1,0,"CBP's objective for full implementation in 2021, a match rate of 97%",97 percent rate objective,"98 percent average technical match rate, August to December 2017",measured,scaled,2026-10-07,not shown,https://www.oig.dhs.gov/sites/default/files/assets/2018-09/OIG-18-80-Sep18.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""U.S. Department of Homeland Security, Office of Inspector General. Progress Made, but CBP Faces Challenges Implementing a Biometric Capability to Track Air Passenger Departures Nationwide. OIG-18-80, 21 September 2018."", ""url"": ""https://www.oig.dhs.gov/sites/default/files/assets/2018-09/OIG-18-80-Sep18.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0675,2017-08-01,event,year,mixed,DHS Inspector General validates that the Traveler Verification Service matched 98% of gate photos in CBP's 2017 exit pilot at nine airports,"The DHS Inspector General validated a 98% average match rate for CBP's face matching at departure gates from August to December 2017, but only 85% of passengers were confirmed.",*Traveler Verification Service* matched 98% of photographed passengers,a 98% average match rate,,found,Found by an auditor,no comparison,very low,aggregate result,false,"Traveler Verification Service (TVS), CBP's facial recognition matching service",,U.S. Customs and Border Protection (the Inspector General does not name the algorithm's supplier),United States,"U.S. Customs and Border Protection, Biometric Entry-Exit Program, with airlines at nine airports (Sprint 8 pilot)",United States,"Atlanta, Boston, Chicago, Houston (Intercontinental and Hobby), Las Vegas, Miami, New York JFK and Washington Dulles airports, United States",United States,Border control (exit records),,Safety and rescue,Images and video,,16,High income,operational_improvement,"Faster data management, decision-making",2,1,3,2,2,5,0.280,organizational,1,0,"CBP's objective for full implementation in 2021, a match rate of 97%",97 percent rate objective,"98 percent average technical match rate, August to December 2017",measured,scaled,2026-10-07,not shown,https://www.oig.dhs.gov/sites/default/files/assets/2018-09/OIG-18-80-Sep18.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""U.S. Department of Homeland Security, Office of Inspector General. Progress Made, but CBP Faces Challenges Implementing a Biometric Capability to Track Air Passenger Departures Nationwide. OIG-18-80, 21 September 2018."", ""url"": ""https://www.oig.dhs.gov/sites/default/files/assets/2018-09/OIG-18-80-Sep18.pdf""}]","{""kind"": ""Official documents""}" AIBD-0835,2017-05-01,event,month,reported,The High Court found that South Wales Police's AFR Locate facial recognition led officers to act in 37 cases where other methods had failed,In September 2019 the High Court found that South Wales Police's live facial recognition had led officers to act in 37 cases. Existing methods had not been able to locate those people.,*AFR Locate* led South Wales Police to act in 37 cases,arrests or disposals in 37 cases,,found,Found by a court,no comparison,very low,aggregate result,false,"AFR Locate, using NEC's NeoFace Watch software",,NEC (now North Gate Public Services (UK) Ltd),Japan,South Wales Police,United Kingdom,"Cardiff and South Wales, United Kingdom",United Kingdom,Policing / public protection,,Safety and rescue,Images and video,Public servants,16,High income,physical_health,Crime reduction or prevention,2,1,4,,3,3,,organizational; societal,1,0,Existing methods of locating wanted people,had not been capable of location by existing methods,arrests or disposals in 37 cases,assumed,litigated,2020-08-11,shown,https://www.naffco.com/downloads/pdf/neoface-watch-brochure.pdf,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""High Court of Justice, Queen's Bench Division, Divisional Court sitting at Cardiff (Haddon-Cave LJ and Swift J). R (Bridges) v Chief Constable of South Wales Police & Ors [2019] EWHC 2341 (Admin), judgment of 4 September 2019. Find Case Law copy."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/admin/2019/2341""}, {""label"": ""Follow-up"", ""value"": ""The Court of Appeal allowed the claimant's appeal on three grounds, including the legal framework for the use. It rejected the ground on proportionality."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewca/civ/2020/1058""}]","{""kind"": ""Legal proceedings""}" AIBD-0713,2017-01-01,event,year,reported,"Sweden's Social Insurance Inspectorate finds that claims picked by Försäkringskassan's risk model had errors in 55% of checks, against 34% at random","Sweden's Social Insurance Inspectorate found that 55% of parental-benefit claims picked by a risk model in 2017 had errors, against 34% of claims checked at random.",Swedish risk model finds more errors in parental-benefit claims,55% of parental-benefit claims picked by a risk model,,found,Found by an auditor,controlled comparison,low,aggregate result,false,"Försäkringskassan's risk-based selection profile (urvalsprofil) for checks of temporary parental benefit, run through the RaKuR infrastructure",,"Försäkringskassan (Swedish Social Insurance Agency), Urvalsanalys unit",Sweden,Försäkringskassan (Swedish Social Insurance Agency),Sweden,Sweden,Sweden,Social insurance control,,Finance,Prediction from data,,16,High income,operational_improvement,Enhanced productivity,2,1,,2,3,,,organizational,2,0,Claims picked at random for checks in 2017,34 procent,55 procent,measured,withdrawn,,shown,https://www.imy.se/globalassets/dokument/beslut/2025/beslut-tillsyn-forsakringskassan.pdf,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Audit report"", ""value"": ""Inspektionen för socialförsäkringen (ISF). Profilering som urvalsmetod för riktade kontroller: En granskning av träffsäkerheten, effektiviteten och rättssäkerheten i Försäkringskassans modeller för riskbaserade kontroller. ISF Rapport 2018:5, March 2018. Chapter 6, printed pp. 109-110, Tabell 7."", ""url"": ""https://isf.se/download/18.6e75aae16a591304896b99/1565330421562/Profilering%20som%20urvalsmetod%20fo%CC%88r%20riktade%20kontroller-ISF-Rapport%202018-05.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Integritetsskyddsmyndigheten (IMY). Beslut efter tillsyn enligt dataskyddsförordningen – Försäkringskassan. IMY-2025-11828, 18 November 2025."", ""url"": ""https://www.imy.se/globalassets/dokument/beslut/2025/beslut-tillsyn-forsakringskassan.pdf""}, {""label"": ""Follow-up"", ""value"": ""Försäkringskassan told the Authority for Privacy Protection that it had taken its machine-learning risk profile for these checks out of use. It did so about a month before its reply of 5 September 2025 and had no plans to bring the profile back."", ""url"": ""https://www.imy.se/globalassets/dokument/beslut/2025/beslut-tillsyn-forsakringskassan.pdf""}]","{""kind"": ""Withdrawals""}" AIBD-0820,2016-01-01,event,month,mixed,New York City's Taxi and Limousine Commission finds that tailgating alerts from Mobileye's system fell in fitted cabs during its safety trial,"In a TLC trial, tailgating alerts from Mobileye's system fell from 100 per 100 miles in July 2015 to between 20 and 40 from January 2016. Crash rates in the trial did not clearly fall.",Tailgating alerts from *Mobileye* fall in New York taxis,fell from 100 per 100 miles in July 2015 to between 20 and 40,,found,Found by a regulator,before and after,very low,aggregate result,false,"Mobileye aftermarket warning system, with a black box recorder added for the trial",,Mobileye,Israel,"A fleet of mostly yellow taxis licensed by the TLC, in the TLC's Vehicle Safety Technology Pilot",United States,"New York City, New York, United States",United States,Road safety / for-hire vehicles,,Safety and rescue,Images and video,Drivers and travelers,11,High income,physical_health,Hazard reduction,2,2,4,,2,5,,personal; societal,1,0,Alert rates in the same fleet early in the trial,100 alerts for each 100 miles traveled (or one alert per mile) in July 2015,between 20 and 40 alerts per 100 miles starting in January 2016,measured,unknown,,shown,https://ims.mobileye.com/fleets/uk/blog/what-does-ai-have-to-do-with-collision-avoidance-systems/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""New York City Taxi and Limousine Commission. Vehicle Safety Technology Final Report. March 2018."", ""url"": ""https://www.nyc.gov/assets/tlc/downloads/pdf/final_vehicle_safety_technology_report.pdf""}, {""label"": ""Regulatory document"", ""value"": ""New York City Taxi and Limousine Commission. Vehicle Safety Technology Final Report. March 2018."", ""url"": ""https://www.nyc.gov/assets/tlc/downloads/pdf/final_vehicle_safety_technology_report.pdf""}]","{""kind"": ""Official documents""}" @@ -102,7 +102,7 @@ AIBD-0817,2026-06-17,source publication,day,reported,"The RDW reports that almos AIBD-1047,2026-06-16,source publication,day,reported,Care Inspectorate Wales and Healthcare Inspectorate Wales find that AI-enabled recording tools accurately record people's own words in Conwy,Care Inspectorate Wales and Healthcare Inspectorate Wales found that Conwy's learning disability practitioners use AI-enabled recording tools to accurately record people's own words.,*AI-enabled recording tools* keep people's own words in Conwy records,,,found,Found by an assessment body,no comparison,very low,program,false,AI-enabled recording tools used by community learning disability practitioners (not named in the letter),,Not named in the letter,,"Conwy County Borough Council and Betsi Cadwaladr University Health Board, Community Learning Disability Team",United Kingdom,"Conwy, Wales, United Kingdom",United Kingdom,Learning disability services (social care and health),,Accessibility,Speech and sound; Text and language,Public servants; People with disabilities,3; 10,High income,operational_improvement,,1,1,,4,3,,,organizational; personal,1,0,,,,,unknown,,not shown,https://www.careinspectorate.wales/sites/default/files/2026-06/260616-conwy-cldt-letter-en.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Care Inspectorate Wales and Healthcare Inspectorate Wales. Joint assurance check letter: Community Learning Disability Team in Conwy (Conwy County Borough Council and Betsi Cadwaladr University Health Board). Assurance check 9 to 11 February 2026, letter published 16 June 2026. Section 3, People, Strengths, paragraph 3.1. The sentence begins 'Practitioners use AI-enabled' on page 2 and ends on page 3. The inspectorates state it as their own finding and do not attribute it to staff."", ""url"": ""https://www.careinspectorate.wales/sites/default/files/2026-06/260616-conwy-cldt-letter-en.pdf""}, {""label"": ""Assessment report"", ""value"": ""Care Inspectorate Wales and Healthcare Inspectorate Wales. Joint assurance check letter: Community Learning Disability Team in Conwy. Section 5, Methodology, paragraph 5.1."", ""url"": ""https://www.careinspectorate.wales/sites/default/files/2026-06/260616-conwy-cldt-letter-en.pdf""}]","{""kind"": ""Official documents""}" AIBD-0706,2026-06-10,event,day,reported,"Brazil's Federal Court of Accounts finds that Galileu's drafts cover 93.3% of claims in labor cases, with hallucinations in 7.3% of topics","Brazil's Federal Court of Accounts tested the Galileu judgment-drafting tool on 50 real cases. Its drafts covered 93.3% of the parties' claims, with hallucinations in 7.3% of topics.",Brazil's audit court finds *Galileu* drafts reliable in labor cases,93.3% of the parties' claims,,found,Found by an auditor,no comparison,very low,aggregate result,false,Galileu (generative AI drafting assistant on Google Gemini Flash models),,"Tribunal Regional do Trabalho da 4ª Região (TRT-4), using Google Gemini models through the cloud contract of the Conselho Superior da Justiça do Trabalho (CSJT)",Brazil,"TRT-4 (Rio Grande do Sul), with a national rollout to the Labor Justice coordinated by the CSJT",Brazil,"Rio Grande do Sul, Brazil",Brazil,Justice administration: labor court drafting,,Public sector,Text and language,Public servants,16,Upper middle income,operational_improvement,Enhanced productivity,1,3,4,3,3,,,organizational,1,0,,,,,scaled,,shown,https://contas.tcu.gov.br/sagas/SvlVisualizarRelVotoAc?codFiltro=SAGAS-SESSAO-ENCERRADA&seOcultaPagina=S&item0=919995,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Tribunal de Contas da União. Acórdão 2836/2026 – TCU – Plenário (TC 001.458/2026-4), relator Ministro Odair Cunha, session of 14 October 2026. Operational audit of the Galileu system at TRT-4, report by the Unidade de Auditoria Especializada em Tecnologia da Informação (AudTI), items 2.2, 2.6 and 155."", ""url"": ""https://contas.tcu.gov.br/sagas/SvlVisualizarRelVotoAc?codFiltro=SAGAS-SESSAO-ENCERRADA&seOcultaPagina=S&item0=919995""}]","{""kind"": ""Official documents""}" AIBD-1165,2026-03-01,source publication,month,reported,UNICEF's Evaluation Office finds that AI analysis of complaint narratives gave insights that appeared specific enough for UNICEF programs to use,"UNICEF's Evaluation Office found that two UNICEF teams used AI to analyze narratives of feedback, including complaints. The analysis gave insights that appeared specific enough for programs to use.",AI analysis of complaint narratives yields trends for UNICEF programs,,,found,Found by an evaluation office,no comparison,very low,benefit created,false,AI analysis of feedback and complaint narratives used by UNICEF's GRM and SBC teams (unnamed tool),,Not named in the report,,"UNICEF's grievance redress mechanism (GRM) and social and behavior change (SBC) teams, in a country office that the passage does not name (the report's GRM and SBC examples are from UNICEF Afghanistan)",,Country office not named in the passage (the report's GRM and SBC examples are from Afghanistan),Afghanistan,Humanitarian accountability and community feedback,,Humanitarian,Text and language,Public servants,16,Low income,operational_improvement,"Faster data management, decision-making",1,1,,,3,,,organizational,1,0,,,,,unknown,,not shown,https://evaluationreports.unicef.org/GetDocument?documentID=20750&fileID=60839,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""UNICEF Evaluation Office. Evaluation of UNICEF's Accountability to Affected Populations: Evaluation report. Carried out by Itad. New York, March 2026. The evaluation covers 2020 to 2025."", ""url"": ""https://evaluationreports.unicef.org/GetDocument?documentID=20750&fileID=60839""}]","{""kind"": ""Official documents""}" -AIBD-0891,2026-01-01,source publication,month,mixed,"The Austrian Institute for Health Technology Assessment found that clinicians felt less burdened by documentation with AI scribes, but time savings varied","In a review of seven reviews covering 36 studies, AIHTA found that clinicians were generally satisfied with AI scribes. They felt less burdened by documentation. Time savings were inconsistent.",Clinicians feel less burdened by documentation with *AI scribes*,from a 19-92% decrease to a 13-50% increase,,found,Found by an assessment body,,very low,aggregate result,false,"AI scribes (ambient scribes), including Dragon Ambient eXperience, Abridge, Nabla Copilot, Sunoh.ai and Amazon HealthScribe",,"Several makers, including Nuance (Dragon Ambient eXperience), Abridge, Nabla, Sunoh.ai and Amazon","Multiple: United States, France","Hospitals, clinics and other services in the studies that AIHTA reviewed",Multiple,"Multiple countries (studies), mostly the United States",United States,Healthcare (clinical documentation),,Health,Speech and sound; Text and language,Health workers,3,High income,operational_improvement,Enhanced productivity,1,,,,2,,,personal; organizational,2,0,"Typing and dictation, in one randomized trial",,"2.7x faster than typing, 2x faster than dictation for patient history sections",measured,persisted,2026-05-01,shown,https://eprints.aihta.at/1597/1/HTA-Projektbericht_Nr.171a.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Erdös J, Grabenhofer L. Artificial Intelligence for Hospital Documentation Support: A Scoping Review of Current Use Cases. AIHTA Project Report No. 171a. Vienna: Austrian Institute for Health Technology Assessment, January 2026. Executive summary and Table 3-3."", ""url"": ""https://eprints.aihta.at/1597/1/HTA-Projektbericht_Nr.171a.pdf""}, {""label"": ""Assessment report"", ""value"": ""Centre for Assessment of Medical Technology in Örebro (CAMTÖ), Region Örebro län (2026). A Systematic Review Comparing Ambient Scribes and Conventional Documentation Methods in Outpatient Care (HTA-rapport 2026:86), Abstract, Conclusion (p. 5)."", ""url"": ""https://www.regionorebrolan.se/siteassets/media/forskning/hta-camto/rapporter/rapporter-2026/2026.86-a-systematic-review-comparing-ambient-scribes-and-conventional-documentation-methods-in-outpatient-care.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0891,2026-01-01,source publication,month,mixed,"The Austrian Institute for Health Technology Assessment found that clinicians felt less burdened by documentation with AI scribes, but time savings varied","In a review of seven reviews covering 36 studies, AIHTA found that clinicians were generally satisfied with AI scribes. They felt less burdened by documentation. Time savings were inconsistent.",Clinicians feel less burdened by documentation with *AI scribes*,from a 19-92% decrease to a 13-50% increase,,found,Found by an assessment body,no comparison,very low,aggregate result,false,"AI scribes (ambient scribes), including Dragon Ambient eXperience, Abridge, Nabla Copilot, Sunoh.ai and Amazon HealthScribe",,"Several makers, including Nuance (Dragon Ambient eXperience), Abridge, Nabla, Sunoh.ai and Amazon","Multiple: United States, France","Hospitals, clinics and other services in the studies that AIHTA reviewed",Multiple,"Multiple countries (studies), mostly the United States",United States,Healthcare (clinical documentation),,Health,Speech and sound; Text and language,Health workers,3,High income,operational_improvement,Enhanced productivity,1,,,,2,,,personal; organizational,2,0,"Typing and dictation, in one randomized trial",,"2.7x faster than typing, 2x faster than dictation for patient history sections",measured,persisted,2026-05-01,shown,https://eprints.aihta.at/1597/1/HTA-Projektbericht_Nr.171a.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Erdös J, Grabenhofer L. Artificial Intelligence for Hospital Documentation Support: A Scoping Review of Current Use Cases. AIHTA Project Report No. 171a. Vienna: Austrian Institute for Health Technology Assessment, January 2026. Executive summary and Table 3-3."", ""url"": ""https://eprints.aihta.at/1597/1/HTA-Projektbericht_Nr.171a.pdf""}, {""label"": ""Assessment report"", ""value"": ""Centre for Assessment of Medical Technology in Örebro (CAMTÖ), Region Örebro län (2026). A Systematic Review Comparing Ambient Scribes and Conventional Documentation Methods in Outpatient Care (HTA-rapport 2026:86), Abstract, Conclusion (p. 5)."", ""url"": ""https://www.regionorebrolan.se/siteassets/media/forskning/hta-camto/rapporter/rapporter-2026/2026.86-a-systematic-review-comparing-ambient-scribes-and-conventional-documentation-methods-in-outpatient-care.pdf""}]","{""kind"": ""Official documents""}" AIBD-1179,2026-01-01,event,year,mixed,New Zealand's Education Review Office finds that teachers using generative AI report less time tailoring resources,"In ERO's 2026 survey, 61 percent of New Zealand secondary teachers report less time tailoring resources. For planning lessons, 44 percent do. Checking student work often offsets these savings.",New Zealand teachers say *generative AI* saves time tailoring resources,61 percent of New Zealand secondary teachers report less time tailoring resources,,found,Found by an auditor,no comparison,very low,aggregate result,false,Generative AI tools used by teachers (the report names no product),,,,Teachers in New Zealand English-medium state and state-integrated schools with students in Year 5 and above,New Zealand,New Zealand (national),New Zealand,Education,,Education,Text and language,Teachers,4,High income,operational_improvement,Enhanced productivity,1,3,1,2,2,,,personal; organizational,1,0,,,,,unknown,,,,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Education Review Office (Te Tari Arotake Mātauranga), Education Evaluation Centre (Te Ihuwaka). Ready or not: How are schools responding to Artificial Intelligence? Research report, 2026."", ""url"": ""https://www.evidence.ero.govt.nz/media/bz2b144q/ready-or-not-how-are-schools-responding-to-artificial-intelligence-research-report.pdf""}, {""label"": ""Audit report"", ""value"": ""Education Review Office (Te Tari Arotake Mātauranga), Education Evaluation Centre (Te Ihuwaka). Ready or not: How are schools responding to Artificial Intelligence? Research report, 2026. Section on primary schools."", ""url"": ""https://www.evidence.ero.govt.nz/media/bz2b144q/ready-or-not-how-are-schools-responding-to-artificial-intelligence-research-report.pdf""}, {""label"": ""Audit report"", ""value"": ""Education Review Office (Te Tari Arotake Mātauranga), Education Evaluation Centre (Te Ihuwaka). Ready or not: How are schools responding to Artificial Intelligence? Research report, 2026. Section on secondary schools."", ""url"": ""https://www.evidence.ero.govt.nz/media/bz2b144q/ready-or-not-how-are-schools-responding-to-artificial-intelligence-research-report.pdf""}]","{""kind"": ""Official documents""}" AIBD-0183,2025-12-16,source publication,day,mixed,Data mining grows from a third to half of French tax controls. The Cour des comptes finds efficiency gains. It also finds a 13.8% share of recovered tax,"The Cour des comptes found that data mining, with machine learning in the tax administration's CFVR program, came to start half of tax controls and made their programming more efficient.","Data mining grows to half of French tax controls, with efficiency gains",originated half of tax controls but accounted for 13.8% of the tax and penalties recovered in 2023,,found,Found by an auditor,before and after,very low,aggregate result,false,"DGFiP Ciblage de la fraude et valorisations des requêtes (CFVR), including machine-learning models for tax-risk analysis",,Direction générale des Finances publiques (DGFiP),France,Direction générale des Finances publiques (DGFiP),France,France (national),France,Tax compliance,Audit targeting and fraud detection,Finance,Prediction from data,Public servants,16,High income,operational_improvement; privacy_protection,Enhanced productivity,1,1,4,,3,,,organizational; societal,5,1,tax controls initiated by DGFiP agents' own analysis,,,measured,scaled,2026-07-21,shown,https://www.ccomptes.fr/sites/default/files/2024-10/20241022-S2024-1165-L-intelligence-artificielle-dans-les-poltiques-publiques-exemple-du-MEF.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes, 'La détection de la fraude fiscale des particuliers', 15 November 2023"", ""url"": ""https://www.ccomptes.fr/fr/publications/la-detection-de-la-fraude-fiscale-des-particuliers""}, {""label"": ""Audit report"", ""value"": ""Cour des comptes, “La lutte contre la fraude fiscale”, 16 December 2025. The French public audit court has no stake in the DGFiP systems. At the end of a cycle of audits it found that mass data cross-matching originates half of tax controls. By its count, cross-matching accounted for 13.8% of the tax and penalties recovered in 2023."", ""url"": ""https://www.ccomptes.fr/fr/publications/la-lutte-contre-la-fraude-fiscale""}, {""label"": ""Audit report"", ""value"": ""Cour des comptes. L’intelligence artificielle dans les politiques publiques : l’exemple du ministère de l’économie et des finances. Observations définitives S2024-1165, 22 October 2024, p. 34. France's audit court has no stake in the tax administration's systems."", ""url"": ""https://www.ccomptes.fr/sites/default/files/2024-10/20241022-S2024-1165-L-intelligence-artificielle-dans-les-poltiques-publiques-exemple-du-MEF.pdf""}]","{""kind"": ""Official documents""}" AIBD-0991,2025-07-07,source publication,day,mixed,The Supreme Court of British Columbia refused to make Acciona redo its machine-learning review of about 7 million documents,In July 2025 a British Columbia court refused to make Acciona redo its TAR 1.0 review of about 7 million documents. It found that a TAR 2.0 review would have cost more than $10.6 million.,BC court declines to make Acciona redo its *TAR 1.0* review,about 7 million documents,,found,Found by a court,no comparison,very low,benefit created,false,"TAR 1.0 predictive coding models (one English, one Spanish)",,"TransPerfect Legal, which built and ran the models (the judgment does not name the TAR software)",United States,"Acciona Wastewater Solutions LP, with its counsel Dentons",Canada,"Vancouver, British Columbia, Canada",Canada,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,1,1,1,,2,1,,organizational,1,0,"A TAR 2.0 (continuous active learning) review, with more manual review",more than $10.6 million,,assumed,unknown,,shown,https://web.archive.org/web/20250806185330id_/https://www.bccourts.ca/jdb-txt/sc/25/12/2025BCSC1256.htm,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Supreme Court of British Columbia (Justice Elwood). Acciona Wastewater Solutions LP v. Greater Vancouver Sewerage and Drainage District, 2025 BCSC 1256, Reasons for Judgment, 7 July 2025, paras 13-16, 24-36 and 96-103. BC Courts text, read through the Internet Archive."", ""url"": ""https://web.archive.org/web/20250806185330id_/https://www.bccourts.ca/jdb-txt/sc/25/12/2025BCSC1256.htm""}]","{""kind"": ""Official documents""}" @@ -117,7 +117,7 @@ AIBD-0229,2024-01-25,event,day,reported,Pindrop's deepfake detector traces a fak AIBD-0677,2024-01-01,event,year,reported,Queensland Audit Office reports that the MPST camera AI cut the photo assessments needing human review by 98.7% in 2024,"The Queensland Audit Office reports that in 2024 the MPST image-recognition AI cut the assessments needing human review at its vendor by 98.7%, to 2.7 million.",*Mobile Phone and Seatbelt Technology* screens out most camera photos,cut the assessments needing human review at its vendor by 98.7%,,found,Found by an auditor,no comparison,very low,aggregate result,false,Mobile Phone and Seatbelt Technology (MPST) image-recognition AI,,"Acusensus, the external vendor that owns and manages the AI (named as TMR's contractor in a ministerial answer, while the audit names no vendor)",Australia,"Queensland Department of Transport and Main Roads, with the Queensland Revenue Office reviewing possible offenses",Australia,"Queensland, Australia",Australia,Traffic enforcement (road safety),,Safety and rescue,Images and video,,11,High income,operational_improvement,Enhanced productivity,1,1,4,,3,5,,organizational; societal,2,0,Human review of every assessment that the AI made,208.4 million assessments performed by the AI in 2024,2.7 million assessments reviewed by the external vendor,assumed,persisted,2026-07-27,shown,https://web.archive.org/web/20260713131003/https://www.qao.qld.gov.au/reports-resources/reports-parliament/managing-ethical-risks-artificial-intelligence,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Queensland Audit Office. Managing the ethical risks of artificial intelligence. Report 2: 2025–26, tabled 24 September 2025, chapter 5 (archived copy of the report page)."", ""url"": ""https://web.archive.org/web/20260713131003/https://www.qao.qld.gov.au/reports-resources/reports-parliament/managing-ethical-risks-artificial-intelligence""}]","{""kind"": ""Official documents""}" AIBD-0689,2024-01-01,event,year,reported,Cour des comptes finds that Panoptes document recognition in France Travail's Upload simplifié avoided about €5 million of contractor costs,The Cour des comptes estimates that automatic document recognition in France Travail's Upload simplifié avoided about €5 million of contractor costs in 2024 and in 2025.,*Upload simplifié* document recognition cuts France Travail's costs,avoided about €5 million of contractor costs in 2024 and in 2025,,found,Found by an auditor,no comparison,very low,aggregate result,false,"Upload simplifié, built on Panoptes (deep-learning computer vision)",,France Travail,France,France Travail,France,France (national),France,Public employment services (document processing),,Public sector,Images and video,,16,High income,operational_improvement,Enhanced productivity,1,1,4,,3,5,,organizational,1,0,Contractors recognizing the types of uploaded documents,,,assumed,unknown,,shown,https://www.ccomptes.fr/sites/default/files/2026-01/20260108-S2025-1558-France-Travail-et-intelligence-artificielle_0.pdf,environ 5 M€,Contractor costs avoided in 2024 and in 2025 through automatic document recognition,True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes. France Travail et l’intelligence artificielle. Observations définitives S2025-1558, adopted 15 October 2025 and published 8 January 2026, p. 58 and annex 5, p. 115."", ""url"": ""https://www.ccomptes.fr/sites/default/files/2026-01/20260108-S2025-1558-France-Travail-et-intelligence-artificielle_0.pdf""}]","{""kind"": ""Official documents""}" AIBD-1038,2023-12-01,source publication,month,mixed,FATF and GAFILAT assessors find that COAF's AI predictive model allows rapid analysis of suspicious transaction reports but sends only 5% to analysts,"FATF and GAFILAT assessors found that COAF's AI predictive model allows rapid analysis of the 4.8 million suspicious transaction reports received in 2018–2022, but that only 5% go to analysts.",*COAF predictive model* speeds triage of Brazil's suspicious reports,4.8 million suspicious transaction reports,,found,Found by an assessment body,no comparison,very low,aggregate result,false,COAF predictive model for scoring suspicious transaction reports,,Conselho de Controle de Atividades Financeiras (COAF) (the report names no other developer),Brazil,"Conselho de Controle de Atividades Financeiras (COAF), Brazil's financial intelligence unit",Brazil,Brazil (national),Brazil,Anti-money laundering: financial intelligence,,Finance,Prediction from data,Public servants,16,Upper middle income,operational_improvement,"Faster data management, decision-making",1,1,4,,2,5,,organizational,1,0,,,,,persisted,2025-10-21,shown,https://web.archive.org/web/20250903135451id_/https://www.fatf-gafi.org/content/dam/fatf-gafi/mer/Brazil-Mutual-Evaluation-2023.pdf.coredownload.inline.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""FATF and GAFILAT. Anti-money laundering and counter-terrorist financing measures – Brazil. Fourth Round Mutual Evaluation Report, December 2023, adopted at the FATF Plenary of October 2023. Chapter 3, Immediate Outcome 6, paras 180-184 and 196-198."", ""url"": ""https://web.archive.org/web/20250903135451id_/https://www.fatf-gafi.org/content/dam/fatf-gafi/mer/Brazil-Mutual-Evaluation-2023.pdf.coredownload.inline.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0654,2023-09-27,source publication,day,reported,NICE allows nine systems in the NHS after concluding that AI autocontouring is quicker than manual contouring for radiotherapy planning,"NICE reviewed 79 studies of clinician-reviewed AI autocontouring by systems such as MVision and Limbus Contour. It found reviewed autocontouring likely equivalent to manual contouring, and quicker.",NICE finds AI outlining quicker for radiotherapy planning,Reported time savings ranged from 3 to 80 minutes,,found,Found by an assessment body,,very low,aggregate result,false,"AI autocontouring software: AI-Rad Companion Organs RT, ART-Plan, DLCExpert, INTContour, Limbus Contour, MIM Contour ProtégéAI, MRCAT Prostate plus Auto-contouring, MVision Segmentation Service and RayStation",,"Siemens Healthineers, TheraPanacea, Mirada Medical, Carina Medical, Limbus AI, MIM Software, Philips, MVision AI and RaySearch","Multiple: Germany, France, United Kingdom, United States, Canada, Netherlands, Finland, Sweden",Radiotherapy centers in the studies reviewed by NICE; NHS radiotherapy services in England,Multiple,"Multiple countries (studies); England, United Kingdom (guidance)",United Kingdom,Healthcare,,Health,Images and video,Health workers,3,High income,operational_improvement,Enhanced productivity,1,2,1,,3,,,organizational,3,0,Manual contouring by radiotherapy staff,Manual contouring time,3 to 80 minutes saved in studies; experts estimated 10 to 30 minutes,measured,persisted,2026-05-15,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC13389072/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI) technologies to aid contouring for radiotherapy treatment planning: early value assessment. HealthTech guidance HTG695 (formerly HTE11), section 3: Committee discussion. 27 September 2023."", ""url"": ""https://www.nice.org.uk/guidance/htg695/chapter/3-Committee-discussion""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HTG695, section 1: Recommendations. 27 September 2023."", ""url"": ""https://www.nice.org.uk/guidance/htg695/chapter/1-Recommendations""}, {""label"": ""Evaluation report"", ""value"": ""King's Technology Evaluation Centre (KiTEC), King's College London, for NHS England's AI in Health and Care Award (2026). A multi-centre real-world evaluation of AI-assisted organ at risk contouring on radiotherapy treatment planning workflows (British Journal of Radiology 2026), Abstract, Results and Conclusions."", ""url"": ""https://pmc.ncbi.nlm.nih.gov/articles/PMC13389072/""}]","{""kind"": ""Official documents""}" +AIBD-0654,2023-09-27,source publication,day,reported,NICE allows nine systems in the NHS after concluding that AI autocontouring is quicker than manual contouring for radiotherapy planning,"NICE reviewed 79 studies of clinician-reviewed AI autocontouring by systems such as MVision and Limbus Contour. It found reviewed autocontouring likely equivalent to manual contouring, and quicker.",NICE finds AI outlining quicker for radiotherapy planning,Reported time savings ranged from 3 to 80 minutes,,found,Found by an assessment body,controlled comparison,low,aggregate result,false,"AI autocontouring software: AI-Rad Companion Organs RT, ART-Plan, DLCExpert, INTContour, Limbus Contour, MIM Contour ProtégéAI, MRCAT Prostate plus Auto-contouring, MVision Segmentation Service and RayStation",,"Siemens Healthineers, TheraPanacea, Mirada Medical, Carina Medical, Limbus AI, MIM Software, Philips, MVision AI and RaySearch","Multiple: Germany, France, United Kingdom, United States, Canada, Netherlands, Finland, Sweden",Radiotherapy centers in the studies reviewed by NICE; NHS radiotherapy services in England,Multiple,"Multiple countries (studies); England, United Kingdom (guidance)",United Kingdom,Healthcare,,Health,Images and video,Health workers,3,High income,operational_improvement,Enhanced productivity,1,2,1,,3,,,organizational,3,0,Manual contouring by radiotherapy staff,Manual contouring time,3 to 80 minutes saved in studies; experts estimated 10 to 30 minutes,measured,persisted,2026-05-15,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC13389072/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI) technologies to aid contouring for radiotherapy treatment planning: early value assessment. HealthTech guidance HTG695 (formerly HTE11), section 3: Committee discussion. 27 September 2023."", ""url"": ""https://www.nice.org.uk/guidance/htg695/chapter/3-Committee-discussion""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HTG695, section 1: Recommendations. 27 September 2023."", ""url"": ""https://www.nice.org.uk/guidance/htg695/chapter/1-Recommendations""}, {""label"": ""Evaluation report"", ""value"": ""King's Technology Evaluation Centre (KiTEC), King's College London, for NHS England's AI in Health and Care Award (2026). A multi-centre real-world evaluation of AI-assisted organ at risk contouring on radiotherapy treatment planning workflows (British Journal of Radiology 2026), Abstract, Results and Conclusions."", ""url"": ""https://pmc.ncbi.nlm.nih.gov/articles/PMC13389072/""}]","{""kind"": ""Official documents""}" AIBD-0883,2023-07-01,event,month,reported,"Together with the police inspectorate, Scotland's Biometrics Commissioner finds that CAID facial matching aided investigations of child abuse","With the police inspectorate, Scotland's Biometrics Commissioner found that CAID facial matching made some contribution to investigations of child abuse. It did the same for the wellbeing of staff.",*CAID* facial matching helps Scottish investigators of child abuse,up to three days to review can now be processed in about an hour,,found,Found by an auditor,no comparison,very low,aggregate result,false,Facial matching function of the UK Child Abuse Image Database (CAID),,"NEC (face recognition technology, per the review), within the Home Office's Child Abuse Image Database, which West Yorkshire Police manages",Japan,"Police Scotland (Cybercrime Investigations, Digital Forensics and the Victim ID Unit)",United Kingdom,"Scotland, United Kingdom",United Kingdom,Policing,,Safety and rescue,Images and video,Public servants; Children,16,High income,operational_improvement; physical_health,Enhanced productivity,1,1,4,,3,4,,organizational; personal; societal,1,0,Manual review of a seized device's images before facial matching,up to three days to review,about an hour,assumed,unknown,,shown,https://www.naffco.com/downloads/pdf/neoface-watch-brochure.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Scottish Biometrics Commissioner and HM Inspectorate of Constabulary in Scotland. Joint Assurance Review of the use of retrospective facial search technologies for criminal justice and police purposes in Scotland (SBC/2025/02). Laid before the Scottish Parliament in March 2025. Summary of findings, section on effectiveness."", ""url"": ""https://www.biometricscommissioner.scot/media/vobnmsze/joint-assurance-review-retrospective-facial-search.pdf""}, {""label"": ""Audit report"", ""value"": ""Scottish Biometrics Commissioner and HM Inspectorate of Constabulary in Scotland. Joint Assurance Review (SBC/2025/02), March 2025, chapter 4, CAID findings. It adds the time saving that operators and managers reported to the reviewers."", ""url"": ""https://www.biometricscommissioner.scot/media/vobnmsze/joint-assurance-review-retrospective-facial-search.pdf""}]","{""kind"": ""Official documents""}" AIBD-0678,2023-06-01,event,month,mixed,"In an audit, the NYC Comptroller finds that police reached ShotSpotter alerts 1 minute 38 seconds faster than 911 reports of outdoor shots in June 2023",New York City's Comptroller found that in June 2023 police reached ShotSpotter alerts 1 minute 38 seconds faster on average than 911 reports of outdoor shots.,Police reached *ShotSpotter* alerts faster than 911 shot reports,1 minute 38 seconds faster,,found,Found by an auditor,controlled comparison,very low,aggregate result,false,ShotSpotter Gunshot Detection and Location System,,"ShotSpotter, Inc. (renamed SoundThinking, Inc. on 10 April 2023)",United States,New York City Police Department (NYPD),United States,"New York City, United States",United States,Policing (gunfire detection),,Safety and rescue,Speech and sound,Public servants,16,High income,physical_health; operational_improvement,Crime reduction or prevention,1,1,1,,2,5,,organizational; societal,1,0,NYPD response to 911 calls about outdoor gunfire in June 2023,5 minutes 28 seconds for 911,3 minutes 50 seconds for ShotSpotter,measured,challenged,2025-02-05,shown,https://www.soundthinking.com/blog/soundthinkings-approach-to-fourth-of-july-safety/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Audit report"", ""value"": ""Office of the New York City Comptroller, Bureau of Audit. Audit Report on the New York City Police Department's Oversight of Its Agreement with ShotSpotter Inc. for the Gunshot Detection and Location System. FP23-074A, 20 June 2024."", ""url"": ""https://comptroller.nyc.gov/reports/audit-report-on-the-new-york-city-police-departments-oversight-of-its-agreement-with-shotspotter-inc-for-the-gunshot-detection-and-location-system/""}, {""label"": ""Follow-up"", ""value"": ""On 5 February 2025 the city Comptroller declined to register NYPD's renewal of the ShotSpotter contract for December 2024 to December 2027, citing violations of procurement rules."", ""url"": ""https://comptroller.nyc.gov/newsroom/comptroller-lander-declines-to-register-shotspotter-contract-renewal""}]","{""kind"": ""Challenges""}" AIBD-1183,2023-06-01,event,month,reported,HMICFRS finds that Bedfordshire Police's automatic redaction technology cuts the time taken to prepare files for court,"In its PEEL inspection published on 29 January 2025, HMICFRS found that Bedfordshire Police's automatic redaction technology improves the time taken to prepare files for court.",*Automatic redaction* speeds court files at Bedfordshire Police,,,found,Found by an auditor,no comparison,very low,benefit created,false,Automatic redaction technology (identified by Police Professional as Riven DocDefender),,"Riven, with Amazon Web Services, according to Police Professional (HMICFRS names no product)",,Bedfordshire Police,United Kingdom,"Bedfordshire, England, United Kingdom",United Kingdom,Policing,,Safety and rescue,Text and language; Images and video,Public servants,16,High income,operational_improvement,Enhanced productivity,1,1,1,,3,,,organizational,2,0,,,,,persisted,2025-03-11,,,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""His Majesty's Inspectorate of Constabulary and Fire & Rescue Services. PEEL 2023-25: An inspection of Bedfordshire Police. 29 January 2025. Original address: https://hmicfrs.justiceinspectorates.gov.uk/peel-reports/bedfordshire-2023-25/ (Wayback copy cited)."", ""url"": ""https://web.archive.org/web/20250517163134id_/https://hmicfrs.justiceinspectorates.gov.uk/peel-reports/bedfordshire-2023-25/""}]","{""kind"": ""Official documents""}" @@ -125,7 +125,7 @@ AIBD-0887,2023-05-01,event,month,mixed,The Scottish Health Technologies Group fo AIBD-0932,2023-01-01,event,year,reported,"In a rapid assessment, IIPH Gandhinagar finds that qXR saved money on reading X-rays from TB screening in India, while Genki was cost-effective","Using 2023 screening records from Tamil Nadu and Maharashtra, researchers costed AI reading at INR 30 per case with qXR. With Genki it was INR 22, against INR 100 for a radiologist.",*qXR* and *Genki* cut the cost of reading X-rays for TB in India,INR 30 with qXR and INR 22 with Genki,,found,Found by an assessment body,before and after,very low,aggregate result,false,qXR and Genki (AI-assisted chest X-ray interpretation for TB screening),,Qure.ai (qXR) and DeepTek (Genki),India,"Public health facilities in Chennai, Tamil Nadu (Genki) and Maharashtra (qXR) screening for TB under India's National Tuberculosis Elimination Program",India,"Chennai, Tamil Nadu, and Maharashtra, India",India,Healthcare (tuberculosis screening),,Health,Images and video,Patients,3,Lower middle income,operational_improvement; physical_health,Enhanced productivity,1,4,1,4,3,,,organizational,2,0,Digital chest X-ray reading by a radiologist,INR 100 (USD 1.2) per case,INR 30 (USD 0.36) for qXR and INR 22 (USD 0.26) for Genki,measured,unknown,,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC12715009/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Health Technology Assessment in India (HTAIn), Department of Health Research, Ministry of Health and Family Welfare; assessment by its resource hub at the Indian Institute of Public Health Gandhinagar (2024). Health Technology Assessment of AI-Assisted CXR for Interpretation for Tuberculosis: A Rapid Health Technology Assessment (Policy Brief, study 77), Executive Summary; Results; Conclusion (pp. 1-2)."", ""url"": ""https://htain.dhr.gov.in/images/pdf/Policy_Briefs/77.AI_based_cXR_Interpreation_Policy_Brief.pdf""}]","{""kind"": ""Official documents""}" AIBD-0998,2023-01-01,event,year,reported,"The FCC finds that fully automatic phone captions cost providers about $0.60 per minute in 2023, against $1.04 with human assistants","The FCC found that providers' allowable expenses averaged about $0.60 per minute for fully automatic captions in 2023, against $1.04 with assistants. Automatic captions carried 74.6% of minutes.",Automatic phone captions cost $0.44 per minute less than assisted ones,about $0.60 per minute for fully automatic captions in 2023,,found,Found by a regulator,controlled comparison,low,aggregate result,false,"ASR-only (fully automatic) Internet Protocol Captioned Telephone Service (IP CTS), as provided by FCC-certified providers",,The certified IP CTS providers and their speech recognition vendors (not named in the order),,FCC-certified IP CTS providers paid from the Interstate TRS Fund (five were paid for IP CTS in 2023),,United States,United States,Telecommunications relay services / accessibility for people with hearing loss,,Accessibility,Speech and sound,People with disabilities,10,High income,operational_improvement,Enhanced productivity,1,2,4,,3,5,,organizational; societal,1,0,"CA-assisted IP CTS, captioned with communications assistants, in 2023",$1.04 per minute,$0.60 per minute,measured,persisted,2026-06-30,shown,https://docs.fcc.gov/public/attachments/FCC-24-81A1.pdf,$0.44 per minute,Lower average allowable expense per minute of ASR-only IP CTS than of CA-assisted IP CTS in 2023,False,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Federal Communications Commission. Report and Order and Order, FCC 24-81, CG Docket Nos. 22-408, 03-123 and 13-24 (Internet Protocol Captioned Telephone Service compensation). Adopted and released 31 July 2024. Public version, para. 20."", ""url"": ""https://docs.fcc.gov/public/attachments/FCC-24-81A1.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Federal Communications Commission. Report and Order and Order, FCC 24-81, CG Docket Nos. 22-408, 03-123 and 13-24 (Internet Protocol Captioned Telephone Service compensation). Adopted and released 31 July 2024. Public version, n. 58."", ""url"": ""https://docs.fcc.gov/public/attachments/FCC-24-81A1.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Federal Communications Commission. Report and Order and Order, FCC 24-81, CG Docket Nos. 22-408, 03-123 and 13-24 (Internet Protocol Captioned Telephone Service compensation). Adopted and released 31 July 2024. Public version, para. 2."", ""url"": ""https://docs.fcc.gov/public/attachments/FCC-24-81A1.pdf""}]","{""kind"": ""Official documents""}" AIBD-1010,2022-12-01,event,month,mixed,ANAO finds that 64.7% of the classifications made by IP Australia's Patent Auto Classifier were correct on average as of November 2025,"From IP Australia's data, the ANAO found that 64.7% of the Patent Auto Classifier's classifications were correct as of November 2025. Missing ones made up 27.5%. The other 7.8% were incorrect.",*Patent Auto Classifier* gets about two in three classifications right,64.7% of the Patent Auto Classifier's classifications were correct,,found,Found by an auditor,no comparison,very low,aggregate result,false,Patent Auto Classifier (PAC),,IP Australia (built in-house and retrained every year),Australia,IP Australia (patent examination),Australia,Australia (national),Australia,Intellectual property administration (patent examination),,Education,Text and language,Public servants,16,High income,operational_improvement,"Faster data management, decision-making",1,1,4,,2,5,,organizational,1,0,,,,,unknown,,shown,https://web.archive.org/web/20260712121955/https://www.anao.gov.au/work/performance-audit/artificial-intelligence-use-ip-australia,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Australian National Audit Office. Artificial Intelligence Use in IP Australia. Auditor-General Report No. 43 of 2025-26, published 29 June 2026. Paragraphs 3.25 and 3.26 and Figure 3.3 (archived copy of the report page)."", ""url"": ""https://web.archive.org/web/20260712121955/https://www.anao.gov.au/work/performance-audit/artificial-intelligence-use-ip-australia""}]","{""kind"": ""Official documents""}" -AIBD-0994,2022-10-21,source publication,day,reported,A special master for e-discovery in a US court found that Huntsman's TAR review had already met his benchmark for a presumptively reasonable search,"In October 2022 a US special master found that Huntsman's TAR review had met his benchmark for a reasonable search, with 9.1% responsive documents in its last two batches.",Huntsman's *TAR* review meets a special master's benchmark,9.1% responsive documents in its last two batches,,found,Found by a court,,very low,benefit created,false,TAR (technology-assisted review),,Not named in the report,,Huntsman Corporation (defendant),United States,"Pittsburgh, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,1,1,1,,3,1,,organizational,1,0,"The special master's benchmark for a presumptively reasonable search, no more than 10% responsive documents in the last two batches reviewed",no more than 10%,9.1%,measured,litigated,2023-11-07,not shown,https://storage.courtlistener.com/recap/gov.uscourts.pawd.250109/gov.uscourts.pawd.250109.800.0.pdf,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Western District of Pennsylvania, E-Discovery Special Master James C. Francis IV (Ret.). In re Diisocyanates Antitrust Litigation, Master Docket Misc. No. 18-1001, MDL No. 2862, Supplemental Report and Recommendation (ECF No. 800), 21 October 2022, pp. 1-2. CourtListener RECAP copy."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.pawd.250109/gov.uscourts.pawd.250109.800.0.pdf""}, {""label"": ""Follow-up"", ""value"": ""The plaintiffs filed limited objections to the 19 October 2022 report. On 7 November 2023 the district court overruled them, adopted both reports as its opinion and required only Wanhua Chemical (America) and Covestro to resume their TAR review."", ""url"": ""https://www.courtlistener.com/docket/8042096/in-re-diisocyanates-antitrust-litigation/?entry_gte=968&entry_lte=968""}]","{""kind"": ""Legal proceedings""}" +AIBD-0994,2022-10-21,source publication,day,reported,A special master for e-discovery in a US court found that Huntsman's TAR review had already met his benchmark for a presumptively reasonable search,"In October 2022 a US special master found that Huntsman's TAR review had met his benchmark for a reasonable search, with 9.1% responsive documents in its last two batches.",Huntsman's *TAR* review meets a special master's benchmark,9.1% responsive documents in its last two batches,,found,Found by a court,no comparison,very low,benefit created,false,TAR (technology-assisted review),,Not named in the report,,Huntsman Corporation (defendant),United States,"Pittsburgh, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,1,1,1,,3,1,,organizational,1,0,"The special master's benchmark for a presumptively reasonable search, no more than 10% responsive documents in the last two batches reviewed",no more than 10%,9.1%,measured,litigated,2023-11-07,not shown,https://storage.courtlistener.com/recap/gov.uscourts.pawd.250109/gov.uscourts.pawd.250109.800.0.pdf,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Western District of Pennsylvania, E-Discovery Special Master James C. Francis IV (Ret.). In re Diisocyanates Antitrust Litigation, Master Docket Misc. No. 18-1001, MDL No. 2862, Supplemental Report and Recommendation (ECF No. 800), 21 October 2022, pp. 1-2. CourtListener RECAP copy."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.pawd.250109/gov.uscourts.pawd.250109.800.0.pdf""}, {""label"": ""Follow-up"", ""value"": ""The plaintiffs filed limited objections to the 19 October 2022 report. On 7 November 2023 the district court overruled them, adopted both reports as its opinion and required only Wanhua Chemical (America) and Covestro to resume their TAR review."", ""url"": ""https://www.courtlistener.com/docket/8042096/in-re-diisocyanates-antitrust-litigation/?entry_gte=968&entry_lte=968""}]","{""kind"": ""Legal proceedings""}" AIBD-0680,2022-06-01,event,month,reported,"Cour des comptes finds that LLaMandement let teams at France's finance ministry absorb 85,345 parliamentary amendments without extra staff","The Cour des comptes found that the DGFiP's LLaMandement sorts amendments for staff. It summarizes them. It let teams absorb 85,345 amendments without extra staff.",*LLaMandement* summarizes amendments for French officials,"85,345 amendments tabled between June 2022 and January 2024",,found,Found by an auditor,no comparison,very low,aggregate result,false,"LLaMendements (LLaMandement), the DGFiP's tool for sorting and summarizing parliamentary amendments",,Direction générale des Finances publiques (DGFiP),France,DGFiP teams handling parliamentary amendments for the economy and finance ministry,France,France (national),France,Public administration (legislative work),,Public sector,Text and language,Public servants,16,High income,operational_improvement,Enhanced productivity,1,1,4,,3,5,,organizational,2,0,,,,,unknown,,shown,https://arxiv.org/abs/2401.16182,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes. L’intelligence artificielle dans les politiques publiques : l’exemple du ministère de l’économie et des finances. Observations définitives S2024-1165, adopted 18 July 2024 and published 22 October 2024, p. 35."", ""url"": ""https://www.ccomptes.fr/sites/default/files/2024-10/20241022-S2024-1165-L-intelligence-artificielle-dans-les-poltiques-publiques-exemple-du-MEF.pdf""}]","{""kind"": ""Official documents""}" AIBD-0685,2022-01-01,event,year,mixed,"Cour des comptes finds that France Travail's Contact via mail matches 98% of jobseekers' emails to their senders, though its reply tool goes unused",The Cour des comptes found that France Travail's machine-learning email tool matches 98% of jobseekers' emails to their senders. It found that the tool's sorting saves advisers time.,*Contact via mail* sorts jobseekers' emails for France Travail advisers,matches 98% of jobseekers' emails to their senders,,found,Found by an auditor,no comparison,very low,aggregate result,false,Contact via mail (CVM),,"Pôle emploi, now France Travail (Intelligence emploi program)",France,France Travail (advisers and managers nationwide),France,France (national),France,Public employment services,,Public sector,Text and language,Public servants,16,High income,operational_improvement,Enhanced productivity,1,1,4,,2,5,,organizational,1,0,,,,,unknown,,shown,https://www.ccomptes.fr/sites/default/files/2026-01/20260108-S2025-1558-France-Travail-et-intelligence-artificielle_0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes. France Travail et l’intelligence artificielle. Observations définitives S2025-1558, adopted 15 October 2025 and published 8 January 2026, annex 5, p. 114 and p. 24."", ""url"": ""https://www.ccomptes.fr/sites/default/files/2026-01/20260108-S2025-1558-France-Travail-et-intelligence-artificielle_0.pdf""}]","{""kind"": ""Official documents""}" AIBD-0684,2021-11-01,event,month,reported,"Cour des comptes finds that 75% of jobseekers use France Travail's automatic CV analysis, which cuts profile filling from 45 to 15 minutes","The Cour des comptes found that 75% of jobseekers use France Travail's automatic CV analysis to fill in their skills profile, which then takes 15 minutes against 45.",*Analyse automatique de CV* fills in jobseekers' skills profiles,75% of jobseekers use France Travail's automatic CV analysis,,found,Found by an auditor,no comparison,very low,aggregate result,false,Analyse automatique de CV (CV discovery),,"Pôle emploi, now France Travail",France,France Travail (jobseekers' online personal space),France,France (national),France,Public employment services,,Public sector,Text and language,Other workers,8,High income,opportunity_access,Employment access,1,4,1,2,3,1,0.160,personal,1,0,Filling in the skills profile without the tool,45 minutes,15 minutes,assumed,unknown,,not shown,https://www.ccomptes.fr/sites/default/files/2026-01/20260108-S2025-1558-France-Travail-et-intelligence-artificielle_0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes. France Travail et l’intelligence artificielle. Observations définitives S2025-1558, adopted 15 October 2025 and published 8 January 2026, pp. 38-39, footnote 19 on p. 23 and annex 5, p. 113."", ""url"": ""https://www.ccomptes.fr/sites/default/files/2026-01/20260108-S2025-1558-France-Travail-et-intelligence-artificielle_0.pdf""}]","{""kind"": ""Official documents""}" @@ -135,7 +135,7 @@ AIBD-0989,2020-04-22,source publication,day,reported,"The High Court in London d AIBD-0985,2020-04-09,source publication,day,reported,"A Kansas federal court found that it was reasonable for Spirit AeroSystems to stop its Predict review at 85% recall, denying a demand for more documents","In April 2020 a US magistrate judge found that Spirit AeroSystems could reasonably stop its Predict review of 322,000 documents at 85% recall. The judge found that more was not proportional.",Court accepts Spirit's *Predict* review stopped at 85% recall,85% recall,,found,Found by a court,no comparison,very low,benefit created,false,"Predict, a technology-assisted review tool that uses continuous active learning",,"Not named in the orders, which say that Spirit's e-discovery vendor Legility offers the tool",,"Spirit AeroSystems, Inc. (defendant), with its e-discovery vendor Legility and its law firm",United States,"Kansas, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Value above replacement,1,1,1,,3,1,,organizational,1,0,Any other form of review,,,assumed,litigated,2020-06-18,shown,https://storage.courtlistener.com/recap/gov.uscourts.ksd.120882/gov.uscourts.ksd.120882.335.0.pdf,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the District of Kansas (Magistrate Judge Angel D. Mitchell). Lawson v. Spirit AeroSystems, Inc., No. 18-1100-EFM-ADM, Memorandum and Order (Doc. 335), 9 April 2020. CourtListener RECAP copy of the filed order."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.ksd.120882/gov.uscourts.ksd.120882.335.0.pdf""}, {""label"": ""Follow-up"", ""value"": ""On 18 June 2020 the court granted Spirit's motion to shift its TAR expenses, reported at about $600,000, to the plaintiff. It found that review of a largely non-responsive data set had become disproportionate to the needs of the case."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.ksd.120882/gov.uscourts.ksd.120882.372.0.pdf""}]","{""kind"": ""Legal proceedings""}" AIBD-0697,2020-01-01,event,year,reported,"UNDP's evaluation office reports that Ecuador's Catalina chatbot answered over 164,900 COVID-19 chats in four months to relieve a government hotline","UNDP's evaluation office found that the Catalina chatbot, set up to unblock Ecuador's COVID-19 hotline, answered over 164,900 chats between May and August 2020.","*Catalina* chatbot answers over 164,900 COVID-19 chats in Ecuador","over 164,900 chats",,found,Found by an evaluation office,no comparison,very low,program,false,Catalina COVID-19 chatbot on Government of Ecuador webpages,,1MillionBot (One Million Bot S.L.),Spain,"Government of Ecuador (Ministry of Telecommunications and the Information Society, MINTEL) with UNDP, which led the initiative, and other UN agencies",Ecuador,Ecuador (national government websites),Ecuador,Public health / government services,,Public sector,Text and language,The public,3,Upper middle income,operational_improvement,Enhanced productivity,1,4,3,,3,,,personal; organizational,3,0,,,,,unknown,,not shown,https://1millionbot.com/en/chatbot-coronavirus-ecuador/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""UNDP Independent Evaluation Office. Independent Country Programme Evaluation: Ecuador. December 2021, Finding 14 (COVID-19). The evaluation covers UNDP's program in Ecuador from 2019 to June 2021."", ""url"": ""https://www.unevaluation.org/get-download-file/5211""}]","{""kind"": ""Official documents""}" AIBD-1034,2019-01-01,event,year,reported,"European Court of Auditors finds that the Flemish paying agency prevented overpayments with checks by monitoring, using data from Sentinel satellites","The European Court of Auditors found that the paying agency of Flanders, like Malta's, used satellite checks by monitoring to prevent a number of overpayments of EU farm aid.",*Checks by monitoring* help Flanders prevent overpayments of farm aid,,,found,Found by an auditor,no comparison,very low,program,false,"Checks by monitoring of the paying agency of Belgium (Flanders), with an artificial intelligence algorithm trained on Sentinel 1 and 2 image time series",,"The paying agency of Belgium (Flanders), following technical guidelines from the European Commission's Joint Research Centre",Belgium,"The paying agency of Belgium (Flanders), for the basic payment scheme",Belgium,"Flanders, Belgium",Belgium,Agriculture (control of EU farm aid claims),,Agriculture and food,Images and video,,16,High income,privacy_protection; operational_improvement,Fraud detection,1,1,4,,3,5,,organizational,2,0,,,,,scaled,2023-01-01,shown,https://www.eca.europa.eu/Lists/ECADocuments/SR20_04/SR_New_technologies_in_agri-monitoring_EN.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""European Court of Auditors. 2020 annual reports on the implementation of the EU budget for the 2020 financial year, Chapter 6 'Natural resources', paragraph 6.20 and Box 6.3. Published 26 October 2021."", ""url"": ""https://www.eca.europa.eu/Lists/ECADocuments/annualreports-2020/annualreports-2020_EN.pdf""}, {""label"": ""Audit report"", ""value"": ""European Court of Auditors. Special report 04/2020: Using new imaging technologies to monitor the Common Agricultural Policy: steady progress overall, but slower for climate and environment monitoring. Box 1 and Table 3. Published 28 January 2020."", ""url"": ""https://www.eca.europa.eu/Lists/ECADocuments/SR20_04/SR_New_technologies_in_agri-monitoring_EN.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0674,2018-01-01,event,year,mixed,"Most staff surveyed by SSA's Inspector General found Insight useful in reviewing hearing decisions, though 71 reported problems with its flags","About 74% of 302 staff who answered a survey by SSA's Inspector General found Insight useful in reviewing hearing decisions. Still, 71 users reported problems with its flags.",Most staff found *Insight* useful for reviewing hearing decisions,About 74% agreed that it was useful and valuable,,found,Found by an auditor,,very low,aggregate result,false,Insight (SSA quality-review software for disability hearing decisions),,Staff of SSA's Office of Appellate Operations (SSA's Office of Analytics and Improvements took over the program after its release),United States,SSA Office of Appellate Operations and Office of Hearings Operations,United States,United States (national),United States,Social security (disability appeals),,Accessibility,Text and language,Public servants,16,High income,operational_improvement,Enhanced productivity,1,3,4,,2,5,,organizational,1,0,"Cases of appellate analysts who did not use Insight, in the appellate office's own study",21.5 percent of non-Insight users' cases returned for quality issues,12.6 percent for Insight users' cases,measured,persisted,,not shown,https://web.archive.org/web/20251010181932id_/https://oig-files.ssa.gov/audits/full/A-12-18-50353.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Social Security Administration, Office of the Inspector General. The Social Security Administration's Use of Insight Software to Identify Potential Anomalies in Hearing Decisions. Audit report A-12-18-50353, 19 April 2019 (archived copy of the report at oig-files.ssa.gov)."", ""url"": ""https://web.archive.org/web/20251010181932id_/https://oig-files.ssa.gov/audits/full/A-12-18-50353.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0674,2018-01-01,event,year,mixed,"Most staff surveyed by SSA's Inspector General found Insight useful in reviewing hearing decisions, though 71 reported problems with its flags","About 74% of 302 staff who answered a survey by SSA's Inspector General found Insight useful in reviewing hearing decisions. Still, 71 users reported problems with its flags.",Most staff found *Insight* useful for reviewing hearing decisions,About 74% agreed that it was useful and valuable,,found,Found by an auditor,no comparison,very low,aggregate result,false,Insight (SSA quality-review software for disability hearing decisions),,Staff of SSA's Office of Appellate Operations (SSA's Office of Analytics and Improvements took over the program after its release),United States,SSA Office of Appellate Operations and Office of Hearings Operations,United States,United States (national),United States,Social security (disability appeals),,Accessibility,Text and language,Public servants,16,High income,operational_improvement,Enhanced productivity,1,3,4,,2,5,,organizational,1,0,"Cases of appellate analysts who did not use Insight, in the appellate office's own study",21.5 percent of non-Insight users' cases returned for quality issues,12.6 percent for Insight users' cases,measured,persisted,,not shown,https://web.archive.org/web/20251010181932id_/https://oig-files.ssa.gov/audits/full/A-12-18-50353.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Social Security Administration, Office of the Inspector General. The Social Security Administration's Use of Insight Software to Identify Potential Anomalies in Hearing Decisions. Audit report A-12-18-50353, 19 April 2019 (archived copy of the report at oig-files.ssa.gov)."", ""url"": ""https://web.archive.org/web/20251010181932id_/https://oig-files.ssa.gov/audits/full/A-12-18-50353.pdf""}]","{""kind"": ""Official documents""}" AIBD-0986,2017-11-27,source publication,day,reported,"After examining New York City's TAR training in camera, a New York federal court found that the City had trained the system appropriately",In November 2017 a US magistrate judge examined New York City's TAR training in camera. She found that the City had handled the system appropriately. She found its process reliable.,Judge clears New York City's *TAR* system after in camera review,"5 documents wrongly marked non-responsive out of the 100,000 reviewed",,found,Found by a court,no comparison,very low,benefit created,false,"The City of New York's technology assisted review (predictive coding) software, not named in the opinion",,Not named in the opinion,,City of New York (defendant) and its document review team,United States,"New York, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,1,1,1,,3,1,,organizational,1,0,Manual review,,,assumed,unknown,,shown,https://storage.courtlistener.com/recap/gov.uscourts.nysd.444418/gov.uscourts.nysd.444418.217.0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Southern District of New York (Magistrate Judge Katharine H. Parker). Winfield v. City of New York, No. 15-CV-05236 (LTS) (KHP), Opinion and Order (Doc. 217), 27 November 2017. CourtListener RECAP copy of the filed opinion."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.nysd.444418/gov.uscourts.nysd.444418.217.0.pdf""}]","{""kind"": ""Official documents""}" AIBD-0681,2017-01-01,event,year,reported,Cour des comptes finds that the ClaudIA chatbot absorbed a fivefold rise in Chorus Pro user requests without extra staff,"The Cour des comptes found that the ClaudIA chatbot absorbed Chorus Pro user requests without extra staff. Requests rose from 20,000 to 100,000 a month.",*ClaudIA* chatbot absorbs a fivefold rise in e-invoicing questions,"from 20,000 to 100,000 a month",,found,Found by an auditor,no comparison,very low,aggregate result,false,"ClaudIA, the Chorus Pro chatbot",,"Agence pour l'informatique financière de l'État (AIFE), using IBM Watson technology according to a 2018 trade report",France,"AIFE, for suppliers using the Chorus Pro e-invoicing portal",France,France (national),France,Public finance (e-invoicing support),,Finance,Text and language,Consumers,16,High income,operational_improvement,Enhanced productivity,1,1,4,,3,5,,organizational,2,0,,,,,unknown,,shown,https://www.lemondeinformatique.fr/actualites/lire-chorus-pro-active-un-chatbot-sous-ibm-watson-71287.html,"1,3 M€","Annual savings planned under the transformation contract for ClaudIA, of which the AIFE reports 55% achieved",True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Cour des comptes. L’intelligence artificielle dans les politiques publiques : l’exemple du ministère de l’économie et des finances. Observations définitives S2024-1165, adopted 18 July 2024 and published 22 October 2024, p. 35 and table 1, p. 16."", ""url"": ""https://www.ccomptes.fr/sites/default/files/2024-10/20241022-S2024-1165-L-intelligence-artificielle-dans-les-poltiques-publiques-exemple-du-MEF.pdf""}]","{""kind"": ""Official documents""}" AIBD-0834,2016-01-01,event,month,mixed,"The US Tax Court held that predictive coding of about 406,000 documents gave the IRS a complete response, at a precision of 3%","In July 2016 the US Tax Court held that a response to the IRS by predictive coding was complete. The model sorted about 406,000 documents into a production set of about 180,000, at 3% precision.",Tax Court finds a response to the IRS by *predictive coding* complete,"about 406,000 documents into a production set of about 180,000",,found,Found by a court,no comparison,very low,benefit created,false,Predictive coding algorithm,,Not named in the order,,"Dynamo Holdings Limited Partnership and Beekman Vista, Inc. (petitioners)",United States,United States,United States,Legal services / e-discovery,,Science,Text and language,,16,High income,operational_improvement,Enhanced productivity,1,2,1,,2,1,,organizational,1,0,,,,,unknown,,shown,https://www.ediscoverylaw.com/wp-content/uploads/2016/08/Dynamo-II-Order.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States Tax Court (Judge Ronald L. Buch). Dynamo Holdings Limited Partnership v. Commissioner, Docket Nos. 2685-11 and 8393-12, order of 13 July 2016 denying the Commissioner's motion to compel. Copy of the signed order on the eDiscovery Law blog of K&L Gates."", ""url"": ""https://www.ediscoverylaw.com/wp-content/uploads/2016/08/Dynamo-II-Order.pdf""}]","{""kind"": ""Official documents""}" @@ -151,7 +151,7 @@ AIBD-0843,2004-07-01,event,month,reported,EPA finds that SmartCEM-75 software tr AIBD-1036,2004-01-01,event,year,reported,GAO finds that FHA's TOTAL Scorecard gave lenders more consistent underwriting decisions than the scorecards that it replaced,In April 2006 GAO found that FHA's TOTAL Scorecard gave lenders less paperwork than the two scorecards that it replaced. Borrowers had less. Underwriting decisions were more consistent.,*TOTAL Scorecard* makes FHA's underwriting more consistent,,,found,Found by an auditor,no comparison,very low,program,false,Technology Open to Approved Lenders (TOTAL) Scorecard,,"Federal Housing Administration (HUD), with contractor Unicon Research Corporation and subcontractor Fair Isaac Corporation",United States,Federal Housing Administration and its direct endorsement lenders,United States,United States (national),United States,Housing finance: mortgage insurance underwriting,,Finance,Prediction from data,Consumers,10; 11,High income,operational_improvement,Value above replacement,1,,4,4,3,5,,organizational; personal,1,0,"The FHA versions of Freddie Mac's Loan Prospector and Fannie Mae's Desktop Underwriter scorecards, used before 2004","at times the results of their assessments differed, which resulted in the same loan being accepted by one scorecard and referred by the other",TOTAL limits the number of loans that need to be approved manually because it provides consistent automatic underwriting decisions,assumed,persisted,2022-09-27,shown,https://www.govinfo.gov/content/pkg/GAOREPORTS-GAO-06-435/pdf/GAOREPORTS-GAO-06-435.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""U.S. Government Accountability Office. Mortgage Financing: HUD Could Realize Additional Benefits from Its Mortgage Scorecard. GAO-06-435, report to the Chairman, Subcommittee on Housing and Community Opportunity, House Committee on Financial Services, 13 April 2006. Section 'HUD Could Benefit Significantly More from TOTAL', pp. 15-16, and Table 1."", ""url"": ""https://www.govinfo.gov/content/pkg/GAOREPORTS-GAO-06-435/pdf/GAOREPORTS-GAO-06-435.pdf""}]","{""kind"": ""Official documents""}" AIBD-0846,2002-05-01,event,month,reported,EPA approves a software monitor for NOx on Braintree's Potter II Unit 3 after finding that it matched reference measurements on both of its fuels,EPA found that a predictive emission monitor on Braintree's Potter II Unit 3 matched reference NOx measurements on gas and oil. It estimated that it would cost less than half as much as an analyzer.,Braintree's *predictive emission monitor* passes EPA's tests for NOx,cost less than half as much as an analyzer,,found,Found by a regulator,tested against a standard,low,aggregate result,false,Predictive emission monitoring system (PEMS) on Potter II Unit 3,,Not named in the letter,,"Braintree Electric Light Department (BELD), Norton P. Potter II Station Unit 3",United States,"Braintree, Massachusetts, United States",United States,Environment and climate,,Environment,Prediction from data,,,High income,operational_improvement,Value above replacement,1,1,4,,3,5,,organizational,1,0,"Hourly NOx readings from a temporary reference method system over the same hours, judged by EPA's three statistical tests",A linear correlation coefficient (r) above 0.8 required,r = 0.8524 on natural gas and r = 0.9211 on fuel oil,measured,unknown,,shown,https://www.epa.gov/sites/default/files/2015-05/documents/potter.pdf,"$12,950",Estimated yearly saving of the PEMS over an analyzer even if outside contractors run the extra audits,True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Environmental Protection Agency, Clean Air Markets Division. Approval of the Predictive Emission Monitoring System Installed on Unit 3 at Braintree Electric Light Department's Norton P. Potter II Station (ORISPL 01660). Letter to Braintree Electric Light Department, 31 May 2006."", ""url"": ""https://www.epa.gov/sites/default/files/2015-05/documents/potter.pdf""}]","{""kind"": ""Official documents""}" AIBD-1160,2026-10-23,source publication,day,tested,China's CMDE finds that the BMP-10 monitor's sedation index matched sedation scores more often than BIS in the maker's trial,"In the maker's trial of 140 patients that China's CMDE reviewed, the BMP-10 monitor's NSI agreed with sedation scores 90.62% of the time, more often than the BIS index.",*BMP-10* sedation index tracked sedation scores better than BIS,agreed with sedation scores 90.62% of the time,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,"BMP-10 anesthesia depth monitor (麻醉深度监护仪), with its sedation depth index (NSI)",Software release 1 (full version 1.0.0.143),"Sichuan Xinyuan Bioelectronics Technology Co., Ltd. (四川新源生物电子科技有限公司)",China,Sichuan Xinyuan Bioelectronics Technology (registration trial with 140 patients at four clinical institutions),China,Four clinical institutions in China that the report does not name,China,Healthcare,,Health,Prediction from data,Patients,3,Upper middle income,physical_health,,,,,,,,,organizational,1,0,"The BIS index, scored against the same MOAA/S sedation assessments in the trial",Lower agreement of the BIS value with MOAA/S (figure not given),"Higher agreement of the NSI with MOAA/S, a statistically significant difference",measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1792720030064015380.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:麻醉深度监护仪 (technical review report for the registration of an anesthesia depth monitor), acceptance no. CQZ2500601, applicant 四川新源生物电子科技有限公司. Undated report, posted on CMDE's website on 23 October 2026. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1792720030064015380.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:麻醉深度监护仪 (technical review report for the registration of an anesthesia depth monitor), acceptance no. CQZ2500601, applicant 四川新源生物电子科技有限公司. Undated report, posted on CMDE's website on 23 October 2026. Section 三、临床评价概述 (clinical evaluation), CMDE's closing judgment on the clinical evidence. The report has no benefit-risk section."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1792720030064015380.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0990,2026-10-05,source publication,day,intended,A federal court in Arizona found insufficient evidence that a 70-80% recall target for technology assisted review is deficient,In October 2026 a federal court in Arizona declined to require a 90-95% recall for technology assisted review. It found insufficient evidence that the defendants' 70-80% target was deficient.,Arizona court declines to raise a 70-80% recall target for *TAR*,a recall rate of 70-80%,,found,Found by a court,,not rated,benefit created,false,Technology Assisted Review (TAR),,Not named in the order,,J.E.T. Limousines & Transportation LLC and the other defendants,United States,"Phoenix, Arizona, United States",United States,Legal services / e-discovery,,Science,Text and language,,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,,,,,unknown,,not shown,https://storage.courtlistener.com/recap/gov.uscourts.azd.1414720/gov.uscourts.azd.1414720.323.0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the District of Arizona (Magistrate Judge John Z. Boyle). McGhee v. J.E.T. Limousines & Transportation LLC, No. CV-24-03394-PHX-DJH, Order (Doc. 323), dated 2 October 2026 and filed 5 October 2026, pp. 3-4 and 8-9. CourtListener RECAP copy."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.azd.1414720/gov.uscourts.azd.1414720.323.0.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0990,2026-10-05,source publication,day,intended,A federal court in Arizona found insufficient evidence that a 70-80% recall target for technology assisted review is deficient,In October 2026 a federal court in Arizona declined to require a 90-95% recall for technology assisted review. It found insufficient evidence that the defendants' 70-80% target was deficient.,Arizona court declines to raise a 70-80% recall target for *TAR*,a recall rate of 70-80%,,found,Found by a court,,very low,benefit created,false,Technology Assisted Review (TAR),,Not named in the order,,J.E.T. Limousines & Transportation LLC and the other defendants,United States,"Phoenix, Arizona, United States",United States,Legal services / e-discovery,,Science,Text and language,,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,,,,,unknown,,not shown,https://storage.courtlistener.com/recap/gov.uscourts.azd.1414720/gov.uscourts.azd.1414720.323.0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the District of Arizona (Magistrate Judge John Z. Boyle). McGhee v. J.E.T. Limousines & Transportation LLC, No. CV-24-03394-PHX-DJH, Order (Doc. 323), dated 2 October 2026 and filed 5 October 2026, pp. 3-4 and 8-9. CourtListener RECAP copy."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.azd.1414720/gov.uscourts.azd.1414720.323.0.pdf""}]","{""kind"": ""Official documents""}" AIBD-1159,2026-08-12,source publication,day,tested,China's CMDE finds that the GluRing ring estimated blood glucose without a needle within 9.20% of venous values on average in the maker's trial,"In the maker's trial that China's CMDE reviewed, the GluRing ring's needle-free glucose estimates differed from venous values by 9.20% on average, with 99.32% in Clarke zones A and B.",*GluRing* estimated blood glucose without a needle in China's review,differed from venous values by 9.20% on average,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,GluRing and GluRing s non-invasive glucose meter (无创血糖仪),Release 1 of the GluRing apps and server (full version 1.1.0),"Beijing Lepu Diagnostics Technology Co., Ltd. (北京乐普诊断科技股份有限公司)",China,Beijing Lepu Diagnostics Technology (registration trial with 96 subjects at three clinical institutions),China,Three clinical institutions in China that the report does not name,China,Healthcare,,Health,Prediction from data,Patients,3,Upper middle income,physical_health,,,,,,,,,personal,1,0,"Venous blood glucose of the same subjects, measured in pairs as the reference standard",,MARD 9.20% (95% CI 8.65% to 9.74%),measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1786495243657085501.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:无创血糖仪 (technical review report for the registration of a non-invasive blood glucose meter), acceptance no. CQZ2500265, applicant 北京乐普诊断科技股份有限公司. Undated report, posted on CMDE's website on 12 August 2026. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1786495243657085501.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:无创血糖仪 (technical review report for the registration of a non-invasive blood glucose meter), acceptance no. CQZ2500265, applicant 北京乐普诊断科技股份有限公司. Undated report, posted on CMDE's website on 12 August 2026. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1786495243657085501.pdf""}]","{""kind"": ""Official documents""}" AIBD-1031,2026-08-06,source publication,day,tested,Canary Islands assessment service SESCS finds that the G-STRIDE gait sensor may detect fall risk better than the Timed Up and Go test,"SESCS found that in one study, the G-STRIDE gait sensor identified more older people at risk of falling than the Timed Up and Go test. It identified about as many as a gait-speed test.",*G-STRIDE* spots more older people at risk of falls than the TUG test,sensitivity of 77%,,found,Found by an assessment body,tested against a standard,very low,aggregate result,false,G-STRIDE (shoe-mounted inertial sensor with a machine-learning gait algorithm),,Academic consortium of Universidad Rey Juan Carlos and collaborators,Spain,"The Spanish validation study that SESCS used, with 157 people over 70",Spain,Spain,Spain,Healthcare (fall prevention in older people),,Health,Prediction from data,Older people,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,The Timed Up and Go test and a gait-speed test in the same people,"test TUG, que tuvo una sensibilidad de 51% (IC95%: 39, 62)","el dispositivo portátil G-STRIDE obtuvo una sensibilidad de 77% (IC95%: 55, 85)",measured,unknown,,shown,https://sescs.es/wp-content/uploads/2026/08/114_2025_SESCS_Telemonitorizacion_Caidas_DEF_NIPO.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Sistemas de telemonitorización portátiles para la evaluación del riesgo de caídas y la detección de caídas en personas de 65 o más años. Informes de Evaluación de Tecnologías Sanitarias, Ministerio de Sanidad, 2026. Resumen ejecutivo (Resultados, Coste-efectividad y análisis económico)."", ""url"": ""https://sescs.es/wp-content/uploads/2026/08/114_2025_SESCS_Telemonitorizacion_Caidas_DEF_NIPO.pdf""}]","{""kind"": ""Official documents""}" AIBD-2308,2026-05-22,source publication,day,tested,FDA De Novo review of PCWP Analysis Software: a machine learning estimate of heart filling pressure met its accuracy goals against catheter measurements,"In the study that FDA reviewed, Cardiosense's PCWP Analysis Software identified elevated heart filling pressure with 78% sensitivity and 80% specificity against catheter measurements in 119 patients.",*PCWP Analysis Software* gauges heart pressure without a catheter,78%,,found,Found by a regulator,tested against a standard,very low,program,false,PCWP Analysis Software (with the CardioTag sensor),,"Cardiosense, Inc.",United States,Cardiosense (clinical validation study at U.S. centers),United States,United States,United States,Healthcare (cardiology),,Health,Prediction from data,Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,Pulmonary capillary wedge pressure measured with a Swan-Ganz catheter during right heart catheterization,Reference standard,Sensitivity 78% and specificity 80% for a pressure of 18 mmHg or more,measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for PCWP Analysis Software (DEN250057): decision summary. Granted 2026-05-22."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for PCWP Analysis Software (DEN250057): decision summary. Granted 2026-05-22."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for PCWP Analysis Software (DEN250057): decision summary. Granted 2026-05-22."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for PCWP Analysis Software (DEN250057): decision summary. Granted 2026-05-22."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for PCWP Analysis Software (DEN250057): decision summary. Granted 2026-05-22."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for PCWP Analysis Software (DEN250057): decision summary. Granted 2026-05-22."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for PCWP Analysis Software (DEN250057): decision summary. Granted 2026-05-22."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for PCWP Analysis Software (DEN250057): decision summary. Granted 2026-05-22."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250057.pdf""}]","{""kind"": ""Official documents""}" @@ -160,10 +160,10 @@ AIBD-1092,2026-05-08,source publication,day,tested,China's CMDE finds that uAI M AIBD-0799,2026-05-04,source publication,day,tested,"FDA found that ArteraAI Breast's high risk group had an 8.7% risk of metastasis within five years, against 0.9% in its low risk group, in 1,271 patients","In FDA's review of ArteraAI Breast, five-year distant metastasis risk was 8.7% in its high risk group and 0.9% in its low risk group, against 3.6% across 1,271 patients.",*ArteraAI Breast* sorted patients with breast cancer by risk of spread,five-year distant metastasis risk was 8.7% in its high risk group,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,ArteraAI Breast,,"Artera, Inc.",United States,"Artera (retrospective clinical validation on archived slides from 1,271 patients at three US sites)",United States,"Three sites, United States",United States,Healthcare,,Health,Images and video,Health workers; Patients,3,High income,physical_health,Medical diagnosis,,,,3,,,,organizational,2,0,"Five-year distant metastasis risk across all 1,271 patients",3.6%,8.7% in the high risk group and 0.9% in the low risk group,measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/K254115.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. 510(k) substantial equivalence determination decision summary for ArteraAI Breast (K254115). Cleared 2026-05-04."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/K254115.pdf""}]","{""kind"": ""Official documents""}" AIBD-1126,2026-05-01,source publication,month,tested,"SHTG finds that AI-estimated CT-FFR agrees well with invasive FFR, so it could spare some patients invasive testing, though no outcome studies exist","SHTG found that machine-learning CT-FFR agrees well with invasive FFR, with pooled sensitivity of 0.86 and specificity of 0.87 per patient. No study followed patients or costs.","*AI-estimated CT-FFR* agrees well with invasive heart tests, SHTG finds",pooled sensitivity of 0.86 and specificity of 0.87,,found,Found by an assessment body,tested against a standard,very low,aggregate result,false,"Machine-learning CT-FFR software in the meta-analyses that SHTG reviewed: cFFR, DEEPVESSEL-FFR, skCT-FFR and XFFR",cFFR versions 1.4 to 3.2 in one meta-analysis,"Siemens Healthineers (cFFR), Keya Medical (DEEPVESSEL-FFR) and other makers","Multiple: Germany, China","Study sites in the meta-analyses that SHTG reviewed, with NHSScotland cardiology services as prospective users",Multiple,"Multiple countries (studies), with an overview for Scotland, United Kingdom",United Kingdom,Healthcare,,Health,Images and video,Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,personal,2,0,"Invasive FFR measurement, the reference test (per patient, 17 studies)",,Pooled sensitivity 0.86 and specificity 0.87,measured,unknown,,shown,https://admin.shtg.scot/media/udwcvr0v/2026-05-18-ct-ffr-imto-v10.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Scottish Health Technologies Group, Healthcare Improvement Scotland. Innovative Medical Technology Overview: Artificial intelligence (AI) estimated computed tomography fractional flow reserve (CT-FFR). May 2026, Key messages."", ""url"": ""https://admin.shtg.scot/media/udwcvr0v/2026-05-18-ct-ffr-imto-v10.pdf""}, {""label"": ""Assessment report"", ""value"": ""Evaluación de Tecnologías Sanitarias de Andalucía (AETSA). Sistemas de estimación del flujo coronario y de la perfusión miocárdica mediante software de análisis de angiografías. Informe de Evaluación de Tecnologías Sanitarias, Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del Sistema Nacional de Salud. Ministerio de Sanidad, 2025. The quotations are in Spanish."", ""url"": ""https://www.aetsa.org/download/publicaciones/22_2025_AETSA_-Sistema-de-estimacion-flujo-coronario_web.pdf""}, {""label"": ""Assessment report"", ""value"": ""Evaluación de Tecnologías Sanitarias de Andalucía (AETSA). Sistemas de estimación del flujo coronario y de la perfusión miocárdica mediante software de análisis de angiografías. Informe de Evaluación de Tecnologías Sanitarias, Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del Sistema Nacional de Salud. Ministerio de Sanidad, 2025. The quotations are in Spanish."", ""url"": ""https://www.aetsa.org/download/publicaciones/22_2025_AETSA_-Sistema-de-estimacion-flujo-coronario_web.pdf""}]","{""kind"": ""Official documents""}" AIBD-0714,2026-04-22,source publication,day,reported,The Swiss Federal Audit Office finds that COMCO's machine-learning screen for bid rigging in construction gave encouraging results in real cases,"The Swiss Federal Audit Office found that COMCO's machine-learning screen for bid rigging in construction gave concrete, encouraging results in real cases. COMCO spends under CHF 10,000 a year on it.",Swiss auditors find that COMCO's bid-rigging screen works at low cost,,,found,Found by an auditor,no comparison,very low,program,false,"COMCO bid-rigging screen for construction tenders (statistical screens combined with machine learning, run as a script)",,"Secretariat of the Swiss Competition Commission (COMCO/WEKO), with the University of Fribourg from 2019",Switzerland,Swiss Competition Commission (COMCO/WEKO),Switzerland,Switzerland,Switzerland,Competition enforcement / public procurement,,Public sector,Prediction from data,Public servants,16,High income,privacy_protection,Fraud detection,,1,4,,3,2,,organizational,1,0,,,,,unknown,,shown,https://www.efk.admin.ch/wp-content/uploads/publikationen/berichte/wirtschaft_und_verwaltung/informatikprojekte/25407/25407be-comco-rapport-utilisation-application-avec-ia.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Contrôle fédéral des finances / Eidgenössische Finanzkontrolle (CDF/EFK). Audit de l'utilisation d'une application avec intelligence artificielle: Commission de la concurrence (COMCO). CDF-25407, version with responses dated 11 February 2026, published 22 April 2026. Section 2.1, p. 11."", ""url"": ""https://www.efk.admin.ch/wp-content/uploads/publikationen/berichte/wirtschaft_und_verwaltung/informatikprojekte/25407/25407be-comco-rapport-utilisation-application-avec-ia.pdf""}]","{""kind"": ""Official documents""}" -AIBD-1124,2026-04-22,source publication,day,intended,"Region Skåne's priority council backs AI grading of eye photos in diabetes screening, expecting about 1,400 staff hours a year to be freed","Region Skåne's council backed AI grading of eye photos for people with diabetes but no known retinopathy. It judged the AI as precise as manual review. It expected it to free about 1,400 hours a year.",Skåne backs *AI fundus grading* to free hours in diabetic eye screening,"about 1,400 hours a year",,found,Found by an assessment body,,not rated,aggregate result,false,"AI-developed automated image analysis of fundus photographs, including DAIRET (Retmarker) and Aireen in the studies behind the main finding",,"Retmarker (DAIRET), Aireen and other makers, with an in-house model in one study","Multiple: Portugal, Czech Republic, Pakistan","Screening services in the studies that HTA South reviewed, with prospective use at Region Skåne's eye clinics",Multiple,"Multiple countries (studies), with a recommendation for Skåne, Sweden",Sweden,Healthcare,,Health,Images and video,Patients,3,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,"Grading of the same photos by human graders, used as the reference standard",,Sensitivity around 89% and specificity 84% for any retinopathy,measured,unknown,,shown,https://vardgivare.skane.se/siteassets/3.-kompetens-och-utveckling/sakkunniggrupper/hta/rapporter/2025/ai-baserad-bildanalys-vid-screening-for-diabetesretinopati.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Region Skåne, Metod- och prioriteringsrådet. Yttrande: AI-utvecklad automatiserad bildanalys av ögonbottenfotografier vid screening för diabetesretinopati. Dnr 2026-O000003, 22 April 2026. Metod- och prioriteringsrådets ställningstagande."", ""url"": ""https://vardgivare.skane.se/siteassets/3.-kompetens-och-utveckling/sakkunniggrupper/metod-och-prioriteringsradet/yttranden/yttrande-ai-bildanalys-retinopati-2026-04-22_jp.pdf""}, {""label"": ""Assessment report"", ""value"": ""HTA syd (HTA South), Region Skåne. AI-baserad bildanalys vid screening för diabetesretinopati: en systematisk översikt och utvärdering av medicinska, hälsoekonomiska, etiska och organisatoriska aspekter. Rapport 2025:4, Lund, 17 December 2025. English abstract, Key Message."", ""url"": ""https://vardgivare.skane.se/siteassets/3.-kompetens-och-utveckling/sakkunniggrupper/hta/rapporter/2025/ai-baserad-bildanalys-vid-screening-for-diabetesretinopati.pdf""}]","{""kind"": ""Official documents""}" +AIBD-1124,2026-04-22,source publication,day,intended,"Region Skåne's priority council backs AI grading of eye photos in diabetes screening, expecting about 1,400 staff hours a year to be freed","Region Skåne's council backed AI grading of eye photos for people with diabetes but no known retinopathy. It judged the AI as precise as manual review. It expected it to free about 1,400 hours a year.",Skåne backs *AI fundus grading* to free hours in diabetic eye screening,"about 1,400 hours a year",,found,Found by an assessment body,,very low,aggregate result,false,"AI-developed automated image analysis of fundus photographs, including DAIRET (Retmarker) and Aireen in the studies behind the main finding",,"Retmarker (DAIRET), Aireen and other makers, with an in-house model in one study","Multiple: Portugal, Czech Republic, Pakistan","Screening services in the studies that HTA South reviewed, with prospective use at Region Skåne's eye clinics",Multiple,"Multiple countries (studies), with a recommendation for Skåne, Sweden",Sweden,Healthcare,,Health,Images and video,Patients,3,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,"Grading of the same photos by human graders, used as the reference standard",,Sensitivity around 89% and specificity 84% for any retinopathy,measured,unknown,,shown,https://vardgivare.skane.se/siteassets/3.-kompetens-och-utveckling/sakkunniggrupper/hta/rapporter/2025/ai-baserad-bildanalys-vid-screening-for-diabetesretinopati.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Region Skåne, Metod- och prioriteringsrådet. Yttrande: AI-utvecklad automatiserad bildanalys av ögonbottenfotografier vid screening för diabetesretinopati. Dnr 2026-O000003, 22 April 2026. Metod- och prioriteringsrådets ställningstagande."", ""url"": ""https://vardgivare.skane.se/siteassets/3.-kompetens-och-utveckling/sakkunniggrupper/metod-och-prioriteringsradet/yttranden/yttrande-ai-bildanalys-retinopati-2026-04-22_jp.pdf""}, {""label"": ""Assessment report"", ""value"": ""HTA syd (HTA South), Region Skåne. AI-baserad bildanalys vid screening för diabetesretinopati: en systematisk översikt och utvärdering av medicinska, hälsoekonomiska, etiska och organisatoriska aspekter. Rapport 2025:4, Lund, 17 December 2025. English abstract, Key Message."", ""url"": ""https://vardgivare.skane.se/siteassets/3.-kompetens-och-utveckling/sakkunniggrupper/hta/rapporter/2025/ai-baserad-bildanalys-vid-screening-for-diabetesretinopati.pdf""}]","{""kind"": ""Official documents""}" AIBD-0816,2026-04-10,source publication,day,tested,"The RDW approves Tesla's FSD Supervised for the Netherlands after more than a year and a half of tests, judging it safer than other driving aids","After more than a year and a half of tests, the Dutch vehicle authority approved Tesla's FSD Supervised. It judged the system safer than other driving aids because it strictly monitors the driver.",Dutch regulator approves *FSD Supervised* after tests on the road,,,found,Found by a regulator,no comparison,very low,aggregate result,false,FSD Supervised (Full Self Driving Supervised),,Tesla,United States,"Tesla, under an EU type approval valid for now only in the Netherlands",United States,Netherlands,Netherlands,Road transport / driver assistance,,Safety and rescue,Machines that act,Drivers and travelers,11,High income,physical_health,Hazard reduction,,,,,,,,personal; societal,1,0,Other driving aids,,"Safer than other driving aids, in the RDW's judgment",assumed,persisted,2026-06-17,not shown,https://www.rdw.nl/nieuws/2026/toelichting-rdw-op-europese-typegoedkeuring-tesla-met-voorlopige-geldigheid-in-nederland,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""RDW (Rijksdienst voor het Wegverkeer, Netherlands Vehicle Authority). Toelichting RDW op Europese typegoedkeuring Tesla met voorlopige geldigheid in Nederland. News item, 10 April 2026 (in Dutch)."", ""url"": ""https://www.rdw.nl/nieuws/2026/toelichting-rdw-op-europese-typegoedkeuring-tesla-met-voorlopige-geldigheid-in-nederland""}]","{""kind"": ""Official documents""}" AIBD-0741,2026-04-02,source publication,day,tested,"NICE allows ArtiQ.Spiro in NHS primary care while evidence is gathered, citing higher sensitivity for airway diseases such as COPD in a UK trial","NICE allowed ArtiQ.Spiro in NHS primary care while evidence is gathered. In the SPIRO-AID trial, it raised professionals' sensitivity for asthma. The same held for COPD. Specificity was similar.",*ArtiQ.Spiro* raised sensitivity for COPD in a UK trial,,,found,Found by an assessment body,randomized trial,moderate,aggregate result,false,ArtiQ.Spiro,,Clario,United States,SPIRO-AID trial with 133 UK primary care professionals reviewing past spirometry records,United Kingdom,United Kingdom,United Kingdom,Healthcare,,Health,Prediction from data,Health workers,3,High income,physical_health,Medical diagnosis,,3,1,,,1,,organizational; personal,2,0,Primary care professionals reviewing the same kind of records without the software (SPIRO-AID control arm),,sensitivity was higher,measured,unknown,,shown,https://www.hra.nhs.uk/planning-and-improving-research/application-summaries/research-summaries/ai-in-primary-care-spirometry-for-diagnosis-of-lung-disease-april/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Digital technologies for applying algorithms to spirometry to support asthma and COPD diagnosis in primary care and community diagnostic centres: early-use assessment. HealthTech guidance HTG776, section 1: Recommendations. 2 April 2026."", ""url"": ""https://www.nice.org.uk/guidance/htg776/chapter/1-Recommendations""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HealthTech guidance HTG776, section 3: Committee discussion (diagnostic accuracy of initial diagnosis). 2 April 2026."", ""url"": ""https://www.nice.org.uk/guidance/htg776/chapter/3-Committee-discussion""}]","{""kind"": ""Official documents""}" -AIBD-1214,2026-03-25,source publication,day,intended,Japan's Kanto District Transport Bureau approves Macnica's NAVYA EVO shuttle in Hitachiota as a Level 4 automated vehicle that needs no driver,On 25 March 2026 the Kanto District Transport Bureau confirmed that the EVO's automated operation device met the safety standards. It approved the shuttle as a Level 4 vehicle on about 200 m of road.,Macnica's *EVO* shuttle may run with no driver on a Hitachiota road,,,found,Found by a regulator,,not rated,benefit created,false,Automated operation device (自動運行装置) on a NAVYA EVO shuttle,,"Navya Mobility (NAVYA), maker of the EVO shuttle",France,"Macnica, Inc. (株式会社マクニカ), the applicant and operator",Japan,"Hitachiota, Ibaraki Prefecture, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,physical_health; operational_improvement,,,,,,,,,societal,3,0,,,,,unknown,,shown,https://web.archive.org/web/20231109122136id_/https://www.navya.tech/?p=827,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 自動運転車(レベル4)の認可を行いました (Approval of a Level 4 automated vehicle). Press release, 25 March 2026."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000370411.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 自動運転車(レベル4)の認可を行いました (Approval of a Level 4 automated vehicle). Press release, 25 March 2026."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000370411.pdf""}]","{""kind"": ""Official documents""}" +AIBD-1214,2026-03-25,source publication,day,intended,Japan's Kanto District Transport Bureau approves Macnica's NAVYA EVO shuttle in Hitachiota as a Level 4 automated vehicle that needs no driver,On 25 March 2026 the Kanto District Transport Bureau confirmed that the EVO's automated operation device met the safety standards. It approved the shuttle as a Level 4 vehicle on about 200 m of road.,Macnica's *EVO* shuttle may run with no driver on a Hitachiota road,,,found,Found by a regulator,,very low,benefit created,false,Automated operation device (自動運行装置) on a NAVYA EVO shuttle,,"Navya Mobility (NAVYA), maker of the EVO shuttle",France,"Macnica, Inc. (株式会社マクニカ), the applicant and operator",Japan,"Hitachiota, Ibaraki Prefecture, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,physical_health; operational_improvement,,,,,,,,,societal,3,0,,,,,unknown,,shown,https://web.archive.org/web/20231109122136id_/https://www.navya.tech/?p=827,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 自動運転車(レベル4)の認可を行いました (Approval of a Level 4 automated vehicle). Press release, 25 March 2026."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000370411.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 自動運転車(レベル4)の認可を行いました (Approval of a Level 4 automated vehicle). Press release, 25 March 2026."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000370411.pdf""}]","{""kind"": ""Official documents""}" AIBD-0767,2026-03-17,source publication,day,tested,"In FDA's De Novo review, Tyto Insights detected eardrum bulging in otoscope videos more accurately than three non-ENT clinicians","In the study that FDA reviewed, Tyto Insights detected eardrum bulging in 857 otoscope recordings with 0.9231 sensitivity, against 0.7242 for three non-ENT clinicians.",*Tyto Insights* detects bulging eardrums better than non-ENT doctors,a sensitivity of 0.9231,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Tyto Insights for Eardrum Bulging Detection,,Tyto Care Ltd.,Israel,Tyto Care (retrospective validation on 857 otoscope recordings from real-world use),Israel,Recording sites not stated in the decision summary,,Healthcare,,Health,Images and video,Children,3,,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,Two pediatricians and one general practitioner reading the same recordings,"Sensitivity 0.7242, specificity 0.7325","Sensitivity 0.9231, specificity 0.9463",measured,persisted,2026-04-27,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250014.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for Tyto Insights for Eardrum Bulging Detection (DEN250014): decision summary. Granted 2026-03-17."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN250014.pdf""}]","{""kind"": ""Official documents""}" AIBD-1020,2026-03-01,event,month,tested,"CATARC found in C-ICAP tests that Xiaomi's assisted driving on the YU7 scored 97.9% for basic driving assistance, with 100% for car following","In CATARC's C-ICAP tests, published on 4 June 2026, Xiaomi's assisted driving on a YU7 scored 97.9% for basic driving assistance. It scored 100% for car following, as it did for combined control.",*Hyper OS 1.12.0* assisted driving scores 97.9% in basic tests on a YU7,97.9% for basic driving assistance,,found,Found by an assessment body,no comparison,very low,aggregate result,false,"Xiaomi assisted driving in Hyper OS 1.12.0, as fitted to the Xiaomi YU7 (小米YU7) long-range rear-wheel-drive version",Hyper OS 1.12.0 (software version tested),Xiaomi Auto (小米汽车),China,"Xiaomi Auto (小米汽车科技有限公司), factory-fitted on the YU7",China,China,China,Road vehicle crash avoidance / assisted driving,,Safety and rescue,Machines that act,Drivers and travelers,11,Upper middle income,physical_health,Hazard reduction,,,,,,,,personal,1,0,,,,,persisted,2026-08-08,shown,https://tech.ifeng.com/c/8oacqKAfyiF,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""China Automotive Technology and Research Center (CATARC), Automotive Test & Assessment Management Center. C-ICAP result for the 小米YU7 超长续航后驱版 (Xiaomi YU7 long-range rear-wheel drive), protocol version 2024 revised, test year 2026. Result page published 4 June 2026, with CATARC's rating report linked from it."", ""url"": ""https://www.c-ncap.org.cn/evaluation/eefd45c3d383aac4ca0e2dc36f20636a""}]","{""kind"": ""Official documents""}" AIBD-0957,2026-02-13,source publication,day,tested,China's CMDE finds that doctors using uAI-BreastTomoCare detected more breast masses on tomosynthesis in the maker's trial,"In the maker's trial that China's CMDE reviewed, doctors using uAI-BreastTomoCare found 62.17% of breast mass lesions on tomosynthesis, against 50.42% without it.",*uAI-BreastTomoCare* helped doctors find more breast masses,"62.17% of breast mass lesions on tomosynthesis, against 50.42% without it",,found,Found by a regulator,controlled comparison,very low,aggregate result,false,uAI-BreastTomoCare (乳腺X射线断层图像辅助检测软件),Release 1.1 (full version 1.1-1-1-45),"United Imaging Intelligence (Beijing) Co., Ltd. (联影智能医疗科技(北京)有限公司)",China,United Imaging Intelligence (Beijing) (retrospective reader trial on images of 432 subjects from three centers),China,"Three trial centers that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers; Women and mothers,3,Upper middle income,physical_health,Medical diagnosis,,,1,,,1,,organizational,1,0,The same doctors reading the same tomosynthesis cases without the software,50.42% lesion-level sensitivity,62.17% lesion-level sensitivity,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1775030830626008899.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:乳腺X射线断层图像辅助检测软件 (technical review report for the registration of breast X-ray tomosynthesis image detection-aid software), acceptance no. CQZ2500562, applicant 联影智能医疗科技(北京)有限公司. 13 February 2026. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1775030830626008899.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:乳腺X射线断层图像辅助检测软件 (technical review report for the registration of breast X-ray tomosynthesis image detection-aid software), acceptance no. CQZ2500562, applicant 联影智能医疗科技(北京)有限公司. 13 February 2026. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1775030830626008899.pdf""}]","{""kind"": ""Official documents""}" @@ -176,7 +176,7 @@ AIBD-0693,2025-12-01,source publication,month,reported,UNDP's evaluation office AIBD-1158,2025-11-21,source publication,day,tested,China's CMDE finds that HB-AI-T1 met its targets for triaging breast thermal images against senior sonographers in the maker's trial,"In the maker's trial of 743 women that China's CMDE reviewed, HB-AI-T1's triage of breast thermal images met its target for sensitivity against senior sonographers. The same held for specificity.",*HB-AI-T1* met its targets for triage of breast images in China's review,,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,HB-AI-T1 (乳腺热层析影像辅助分诊软件),Release 1,"Wuhan Haobo Technology Co., Ltd. (武汉昊博科技有限公司)",China,Wuhan Haobo Technology (registration trial with 743 women at three clinical institutions),China,Three clinical institutions in China that the report does not name,China,Healthcare,,Health,Images and video,Health workers; Women and mothers,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,"Preset targets for sensitivity and specificity, with a panel of senior sonographers' diagnoses as the reference standard",Preset targets that the report does not state,Sensitivity and specificity reached the targets (no figures given),measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1765437352894073517.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:乳腺热层析影像辅助分诊软件 (technical review report for the registration of breast thermal tomography image triage-aid software), acceptance no. CQZ2400541, applicant 武汉昊博科技有限公司. 21 November 2025. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1765437352894073517.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:乳腺热层析影像辅助分诊软件 (technical review report for the registration of breast thermal tomography image triage-aid software), acceptance no. CQZ2400541, applicant 武汉昊博科技有限公司. 21 November 2025. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1765437352894073517.pdf""}]","{""kind"": ""Official documents""}" AIBD-0959,2025-10-29,source publication,day,tested,"China's CMDE finds that with AI-Rad Companion Prostate MR, doctors diagnosed lesions on prostate MRI slightly better in the maker's trial","In the maker's trial that China's CMDE reviewed, doctors reading prostate MRI with AI-Rad Companion Prostate MR reached an AUROC of 0.7010, against 0.6758 without it.",*AI-Rad Companion Prostate MR* lifted readers' accuracy slightly,"an AUROC of 0.7010, against 0.6758 without it",,found,Found by a regulator,controlled comparison,very low,aggregate result,false,AI-Rad Companion Prostate MR (前列腺癌磁共振图像辅助检测软件),"Release VA11 (full version VA11B), algorithm release VA1 (full version VA10A)","Siemens Healthineers, with 上海西门子医疗器械有限公司 (Siemens Shanghai Medical Equipment Ltd.) as the applicant in China",Germany,"上海西门子医疗器械有限公司 (Siemens Shanghai Medical Equipment Ltd.), the applicant (retrospective reader trial with 12 doctors and 340 subjects)",China,"Trial sites that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,2,1,,,1,,organizational,2,0,Doctors reading the same cases without the software,AUROC 0.6758,AUROC 0.7010,measured,scaled,2025-11-07,shown,https://www.cmde.org.cn/directory/web/cmde/images/1765243681648031434.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:前列腺癌磁共振图像辅助检测软件 (technical review report for the registration of prostate cancer MR image detection-aid software), acceptance no. CQZ2401376, applicant 上海西门子医疗器械有限公司. 29 October 2025. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1765243681648031434.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:前列腺癌磁共振图像辅助检测软件 (technical review report for the registration of prostate cancer MR image detection-aid software), acceptance no. CQZ2401376, applicant 上海西门子医疗器械有限公司. 29 October 2025. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1765243681648031434.pdf""}]","{""kind"": ""Official documents""}" AIBD-0739,2025-10-14,source publication,day,tested,"NICE allows five AI tools, such as HealthVCF, in the NHS after finding that they detect spinal fragility fractures missed in radiology reports","NICE found that AI tools detect moderate to severe vertebral fragility fractures missed in original radiology reports, generally with specificity above 90%. It allowed five in the NHS in 2025.",AI tools find spinal fractures missed in routine scan reports,a specificity above 90%,,found,Found by an assessment body,tested against a standard,very low,aggregate result,false,"AI tools for opportunistic detection of vertebral fragility fractures on CT: BriefCase-Triage, CINA-VCF Quantix, HealthVCF, HealthOST and IB Lab FLAMINGO",,"Aidoc Medical (BriefCase-Triage), Avicenna.AI (CINA-VCF Quantix), Nanox AI (HealthVCF and HealthOST) and IB Lab (IB Lab FLAMINGO)","Multiple: Israel, France, Austria",Hospitals in the studies reviewed by NICE and NHS radiology services in England during the evidence generation period,Multiple,"England, United Kingdom (guidance), with studies in several countries",United Kingdom,Healthcare (radiology and osteoporosis),,Health,Images and video,Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,Original radiology reports (standard care),,Additional moderate to severe fractures detected,measured,persisted,,shown,https://www.nice.org.uk/guidance/htg760/chapter/3-Committee-discussion,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI) technologies to aid opportunistic detection of vertebral fragility fractures: early value assessment. HealthTech guidance HTG760 (formerly HTE34), section 3: Committee discussion. 14 October 2025."", ""url"": ""https://www.nice.org.uk/guidance/htg760/chapter/3-Committee-discussion""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HTG760, section 1: Recommendations. 14 October 2025."", ""url"": ""https://www.nice.org.uk/guidance/htg760/chapter/1-Recommendations""}]","{""kind"": ""Official documents""}" -AIBD-1091,2025-10-09,source publication,day,tested,China's CMDE finds that MIAS-3000 flagged intermediate or later macular degeneration on OCT with 95.5% sensitivity in the maker's trial,"In the maker's trial that China's CMDE reviewed, MIAS-3000 flagged intermediate or later age-related macular degeneration on OCT with 95.5% sensitivity and 92.1% specificity per eye.",*MIAS-3000* flagged macular degeneration on OCT scans in China's review,95.5% sensitivity and 92.1% specificity per eye,,found,Found by a regulator,,very low,aggregate result,false,MIAS-3000 (眼科OCT图像辅助诊断软件),Release V1,"Suzhou Bigvision Medical Technology Co., Ltd. (苏州比格威医疗科技有限公司)",China,Suzhou Bigvision Medical Technology (prospective multicenter registration trial with 908 subjects),China,Trial sites in China that the report does not name,China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,A target of 85% for the lower 95% confidence bounds of sensitivity and specificity,Target of 85% for the lower confidence bounds,"Sensitivity 95.5%, 95%CI(92.5%,97.5%) and specificity 92.1%, 95%CI(89.6%,94.2%) per eye",measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1761699620880017467.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:眼科OCT图像辅助诊断软件 (technical review report for the registration of ophthalmic OCT image diagnosis-aid software), acceptance no. CQZ2400999, applicant 苏州比格威医疗科技有限公司. 9 October 2025. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1761699620880017467.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:眼科OCT图像辅助诊断软件 (technical review report for the registration of ophthalmic OCT image diagnosis-aid software), acceptance no. CQZ2400999, applicant 苏州比格威医疗科技有限公司. 9 October 2025. Section 综合评价意见 (concluding evaluation), CMDE's own conclusion and recommendation to approve."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1761699620880017467.pdf""}]","{""kind"": ""Official documents""}" +AIBD-1091,2025-10-09,source publication,day,tested,China's CMDE finds that MIAS-3000 flagged intermediate or later macular degeneration on OCT with 95.5% sensitivity in the maker's trial,"In the maker's trial that China's CMDE reviewed, MIAS-3000 flagged intermediate or later age-related macular degeneration on OCT with 95.5% sensitivity and 92.1% specificity per eye.",*MIAS-3000* flagged macular degeneration on OCT scans in China's review,95.5% sensitivity and 92.1% specificity per eye,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,MIAS-3000 (眼科OCT图像辅助诊断软件),Release V1,"Suzhou Bigvision Medical Technology Co., Ltd. (苏州比格威医疗科技有限公司)",China,Suzhou Bigvision Medical Technology (prospective multicenter registration trial with 908 subjects),China,Trial sites in China that the report does not name,China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,A target of 85% for the lower 95% confidence bounds of sensitivity and specificity,Target of 85% for the lower confidence bounds,"Sensitivity 95.5%, 95%CI(92.5%,97.5%) and specificity 92.1%, 95%CI(89.6%,94.2%) per eye",measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1761699620880017467.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:眼科OCT图像辅助诊断软件 (technical review report for the registration of ophthalmic OCT image diagnosis-aid software), acceptance no. CQZ2400999, applicant 苏州比格威医疗科技有限公司. 9 October 2025. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1761699620880017467.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:眼科OCT图像辅助诊断软件 (technical review report for the registration of ophthalmic OCT image diagnosis-aid software), acceptance no. CQZ2400999, applicant 苏州比格威医疗科技有限公司. 9 October 2025. Section 综合评价意见 (concluding evaluation), CMDE's own conclusion and recommendation to approve."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1761699620880017467.pdf""}]","{""kind"": ""Official documents""}" AIBD-0633,2025-07-31,source publication,day,tested,FDA De Novo review of ArteraAI Prostate: AI risk categories from biopsy slides separate 10-year metastasis risk in 886 patients,"In FDA's review of ArteraAI Prostate, among 886 patients, 10-year metastasis risk was 28.1% in its high risk category, against 8.1% overall. In its low category it was 3.3%.",*ArteraAI Prostate* sorted prostate cancer patients by risk of spread,10-year metastasis risk was 28.1% in its high risk category,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,ArteraAI Prostate,,"Artera, Inc.",United States,Artera (clinical validation at three US sites),United States,"Three sites, United States",United States,Healthcare,,Health,Images and video,Health workers; Patients,3,High income,physical_health,,,,,2,,,,organizational,2,0,Overall 10-year distant metastasis risk in the validation cohort,8.1% overall,28.1% in the high risk category; 3.3% in the low risk category,measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN240068.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for ArteraAI Prostate (DEN240068): decision summary. Granted 2025-07-31."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN240068.pdf""}]","{""kind"": ""Official documents""}" AIBD-1025,2025-06-26,source publication,day,tested,"CATARC found in its C-ICAP tests that the Onvo L60's assisted driving scored 99.4% for basic driving assistance, with 96.3% for navigation","In C-ICAP results that CATARC published on 26 June 2025, the Onvo L60's assisted driving scored 99.4% for basic driving assistance. It scored 96.3% for navigation pilot assistance.",*Onvo L60 assisted driving* scores 99.4% for basic driving assistance,99.4% for basic driving assistance,,found,Found by an assessment body,no comparison,very low,aggregate result,false,"Assisted driving (辅助驾驶), as fitted to the Onvo (乐道) L60 60kWh rear-wheel drive",1.1.1CN (assisted-driving software version tested),"NIO (蔚来汽车科技(安徽)有限公司), for its Onvo brand",China,NIO (factory-fitted on the Onvo L60),China,China,China,Road vehicle crash avoidance / assisted driving,,Safety and rescue,Machines that act,Drivers and travelers,11,Upper middle income,physical_health,Hazard reduction,,,,,,,,personal,2,0,,,,,persisted,2026-06-11,not shown,https://www.c-ncap.org.cn/evaluation/ecb71725a7c11d895ff271bc2cdab4b8,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""China Automotive Technology and Research Center (CATARC), Automotive Testing & Assessment Management Center (中汽测评). 乐道L60 [Onvo L60 60kWh 后驱], C-ICAP results page, protocol 2024 edition, test year 2025. Published 26 June 2025. Section 辅助驾驶 > 行车辅助 [assisted driving > driving assistance]."", ""url"": ""https://www.c-ncap.org.cn/evaluation/ecb71725a7c11d895ff271bc2cdab4b8""}]","{""kind"": ""Official documents""}" AIBD-1093,2025-06-26,source publication,day,tested,China's CMDE finds that uAI-CT_CerebralPerfusion measured stroke volumes in close agreement with a marketed product in the maker's trial,"In the maker's trial that China's CMDE reviewed, uAI-CT_CerebralPerfusion measured infarct core and penumbra volumes with ICCs of 0.96 and 0.83 against an already-marketed product.",*uAI-CT_CerebralPerfusion* agreed with a marketed tool on stroke scans,ICCs of 0.96 and 0.83,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,uAI-CT_CerebralPerfusion (脑缺血CT灌注图像辅助评估软件),Release 1.1 (full version 1.1-0-0-56),"Shanghai United Imaging Intelligence Co., Ltd. (上海联影智能医疗科技有限公司)",China,Shanghai United Imaging Intelligence (retrospective registration trial with 446 cases at six institutions),China,"Six trial institutions that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,"An already-marketed cerebral ischemia image assessment software of the same kind on the same cases, with a target ICC of 0.75 for the lower 95% confidence bound",Target ICC of 0.75,"ICC 0.96 (95%CI: 0.95, 0.97) for the infarct core and 0.83 (95%CI: 0.79, 0.86) for the penumbra",measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1751851972285059301.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:脑缺血CT灌注图像辅助评估软件 (technical review report for the registration of cerebral ischemia CT perfusion image assessment-aid software), acceptance no. CQZ2401663, applicant 上海联影智能医疗科技有限公司. 26 June 2025. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1751851972285059301.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:脑缺血CT灌注图像辅助评估软件 (technical review report for the registration of cerebral ischemia CT perfusion image assessment-aid software), acceptance no. CQZ2401663, applicant 上海联影智能医疗科技有限公司. 26 June 2025. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1751851972285059301.pdf""}]","{""kind"": ""Official documents""}" @@ -189,7 +189,7 @@ AIBD-0960,2025-03-06,source publication,day,tested,China's CMDE finds that sonog AIBD-0652,2025-02-27,source publication,day,tested,EMA qualifies AIM-NASH after AI-assisted pathologists scored MASH liver biopsies more accurately than pathologists reading alone,"EMA qualified PathAI's AIM-NASH after a 1,481-case validation in which AI-assisted pathologists agreed better with expert panels on lobular inflammation and on hepatocellular ballooning.",Europe's medicines regulator accepts *AIM-NASH* for biopsy scoring,hepatocellular ballooning (0.563 against 0.414),,found,Found by a regulator,controlled comparison,very low,aggregate result,false,AIM-NASH (Artificial Intelligence-Based Measurement of NASH histology; also called AIM-MASH),,PathAI,United States,"PathAI with sponsors of MASH drug trials (validation on archived biopsies from trials of obeticholic acid, pegbelfermin and semaglutide)",United States,Multinational MASH clinical trials (central pathology reading),,Drug development / pathology,,Science,Images and video,Health workers,3,,physical_health,Medical diagnosis,,,,,,,,organizational,4,0,"Individual pathologists reading without the tool, scored against the same expert panel",Mean weighted kappa of unassisted reads,Weighted kappa higher by 0.15 for ballooning and 0.12 for lobular inflammation,measured,scaled,2025-12-08,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC11750710/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""European Medicines Agency, Committee for Medicinal Products for Human Use. Qualification opinion for Artificial Intelligence-Based Measurement of Non-alcoholic Steatohepatitis Histology in Liver Biopsies to Determine Disease Activity in NASH/MASH Clinical Trials. Adopted 27 February 2025 (EMADOC-1700519818-1761332 Corr.1)."", ""url"": ""https://www.ema.europa.eu/en/documents/other/qualification-opinion-artificial-intelligence-based-measurement-non-alcoholic-steatohepatitis-histology-liver-biopsies-determine-disease-activity-nash-mash-clinical-trials_en.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Drug Evaluation and Research. FDA Qualifies First AI Drug Development Tool, Will Be Used in 'MASH' Clinical Trials. 8 December 2025."", ""url"": ""https://www.fda.gov/drugs/drug-safety-and-availability/fda-qualifies-first-ai-drug-development-tool-will-be-used-mash-clinical-trials""}]","{""kind"": ""Official documents""}" AIBD-0961,2025-02-25,source publication,day,tested,China's CMDE finds that less senior readers detected breast lesions more accurately with Yizhun-DBU in the maker's trial,"In the maker's trial that China's CMDE reviewed, six less senior readers detected breast lesions with an AFROC-AUC of 0.8785 using Yizhun-DBU, against 0.6500 alone.",*Yizhun-DBU* helped less senior readers find breast lesions,"an AFROC-AUC of 0.8785 using Yizhun-DBU, against 0.6500 alone",,found,Found by a regulator,controlled comparison,very low,aggregate result,false,Yizhun-DBU (乳腺超声影像辅助检测软件),Release 4,"Beijing Yizhun Medical Technology Co., Ltd. (北京医准医疗科技有限公司)",China,Beijing Yizhun Medical Technology (prospective reader trial with 176 subjects at three sites),China,"Three trial sites that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,2,1,,,1,,organizational,1,0,"The same six readers reading the same cases without the software, with three senior readers as the reference","AFROC-AUC 0.6500(0.6268, 0.6733)","AFROC-AUC 0.8785 (0.8471, 0.9100)",measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1741742380821028657.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:乳腺超声影像辅助检测软件 (technical review report for the registration of breast ultrasound image detection-aid software), acceptance no. CQZ2400392, applicant 北京医准医疗科技有限公司. 25 February 2025. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1741742380821028657.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:乳腺超声影像辅助检测软件 (technical review report for the registration of breast ultrasound image detection-aid software), acceptance no. CQZ2400392, applicant 北京医准医疗科技有限公司. 25 February 2025. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1741742380821028657.pdf""}]","{""kind"": ""Official documents""}" AIBD-1086,2025-02-19,source publication,day,tested,China's CMDE finds that Infervision's IR-CT-PE flagged pulmonary embolism with 92.27% sensitivity at 96.91% specificity in the maker's trial,"In the maker's two-site trial that China's CMDE reviewed, Infervision's IR-CT-PE flagged pulmonary embolism with 92.27% sensitivity and 96.91% specificity, above 80% targets.",*IR-CT-PE* flagged pulmonary embolism in a two-site trial in China,92.27% sensitivity and 96.91% specificity,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,"IR-CT-PE (肺栓塞CT血管造影图像辅助分诊软件), models IR-CT-PE(A) and IR-CT-PE(B)",Release 1 (full version 1.0.3.0),"Infervision Medical Technology Co., Ltd. (推想医疗科技股份有限公司)",China,Infervision Medical Technology (trial with 485 cases at two institutions),China,Two trial institutions in China that the report does not name,China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,"Preset targets of 80%, with an expert panel's case judgments as the reference",Targets of 80% for sensitivity and for specificity,92.27% sensitivity and 96.91% specificity,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1740618990620049428.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:肺栓塞CT血管造影图像辅助分诊软件 (technical review report for the registration of pulmonary embolism CT angiography image triage-aid software), acceptance no. CQZ2400926, applicant 推想医疗科技股份有限公司. 19 February 2025. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1740618990620049428.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:肺栓塞CT血管造影图像辅助分诊软件 (technical review report for the registration of pulmonary embolism CT angiography image triage-aid software), acceptance no. CQZ2400926, applicant 推想医疗科技股份有限公司. 19 February 2025. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1740618990620049428.pdf""}]","{""kind"": ""Official documents""}" -AIBD-1014,2025-02-18,source publication,day,intended,Japan's Kinki District Transport Bureau approves Osaka Metro's Expo park-and-ride bus as a Level 4 automated vehicle that needs no driver,On 18 February 2025 a Japanese transport bureau approved Osaka Metro's Expo park-and-ride bus for Level 4. It confirmed that the bus's automated operation device met the safety standards.,*ASM-AD* lets Osaka Metro run an Expo bus with no driver,,,found,Found by a regulator,,not rated,benefit created,false,Automated operation device (ASM-AD by Advanced Smart Mobility) on an EV Motors Japan F8 series2-City Bus 10.5m,,"Advanced Smart Mobility Co., Ltd. (先進モビリティ株式会社), maker of the ASM-AD system, with the bus built by EV Motors Japan",Japan,"Osaka Metro (大阪市高速電気軌道株式会社), the applicant and operator",Japan,"Maishima, Osaka, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,Drivers and travelers,11,High income,physical_health; operational_improvement,,,,,,,,,societal,2,0,,,,,withdrawn,2025-04-28,not shown,https://jidounten-lab.com/u_52872,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Kinki District Transport Bureau, Automotive Technology and Safety Department. 近畿初!自動運転車(レベル4)認可 ~ 万博来場者向け自動運転を実現 ~ (First in Kinki: Level 4 automated vehicle approved for Expo visitors). News release, 18 February 2025, first paragraph and table."", ""url"": ""https://wwwtb.mlit.go.jp/kinki/content/000343237.pdf""}, {""label"": ""Follow-up"", ""value"": ""After a bus with Advanced Smart Mobility's system hit a wall on 28 April 2025, Osaka Metro stopped running its self-driving Expo buses, blamed a setting error, and planned test runs before resuming."", ""url"": ""https://www.netdenjd.com/articles/-/318720""}]","{""kind"": ""Withdrawals""}" +AIBD-1014,2025-02-18,source publication,day,intended,Japan's Kinki District Transport Bureau approves Osaka Metro's Expo park-and-ride bus as a Level 4 automated vehicle that needs no driver,On 18 February 2025 a Japanese transport bureau approved Osaka Metro's Expo park-and-ride bus for Level 4. It confirmed that the bus's automated operation device met the safety standards.,*ASM-AD* lets Osaka Metro run an Expo bus with no driver,,,found,Found by a regulator,,very low,benefit created,false,Automated operation device (ASM-AD by Advanced Smart Mobility) on an EV Motors Japan F8 series2-City Bus 10.5m,,"Advanced Smart Mobility Co., Ltd. (先進モビリティ株式会社), maker of the ASM-AD system, with the bus built by EV Motors Japan",Japan,"Osaka Metro (大阪市高速電気軌道株式会社), the applicant and operator",Japan,"Maishima, Osaka, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,Drivers and travelers,11,High income,physical_health; operational_improvement,,,,,,,,,societal,2,0,,,,,withdrawn,2025-04-28,not shown,https://jidounten-lab.com/u_52872,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Kinki District Transport Bureau, Automotive Technology and Safety Department. 近畿初!自動運転車(レベル4)認可 ~ 万博来場者向け自動運転を実現 ~ (First in Kinki: Level 4 automated vehicle approved for Expo visitors). News release, 18 February 2025, first paragraph and table."", ""url"": ""https://wwwtb.mlit.go.jp/kinki/content/000343237.pdf""}, {""label"": ""Follow-up"", ""value"": ""After a bus with Advanced Smart Mobility's system hit a wall on 28 April 2025, Osaka Metro stopped running its self-driving Expo buses, blamed a setting error, and planned test runs before resuming."", ""url"": ""https://www.netdenjd.com/articles/-/318720""}]","{""kind"": ""Withdrawals""}" AIBD-0962,2025-02-13,source publication,day,tested,"China's CMDE finds that with DC-OPGAI-01, dentists found more decayed teeth on panoramic X-rays in the maker's trial","In the maker's trial that China's CMDE reviewed, 12 dentists using DC-OPGAI-01 detected caries with a tooth-level sensitivity of 82.40%, against 67.43% without it.",*DC-OPGAI-01* helped dentists find more tooth decay on X-rays,"a tooth-level sensitivity of 82.40%, against 67.43% without it",,found,Found by a regulator,controlled comparison,very low,aggregate result,false,DC-OPGAI-01 (龋病口腔曲面体层图像辅助检测软件),Release 1 (full version 1.0.0),"Beijing Yuyi Ganlan Information Technology Co., Ltd. (北京羽医甘蓝信息技术有限公司)",China,Beijing Yuyi Ganlan Information Technology (retrospective reader trial with 12 dentists and 402 subjects at three sites),China,"Three trial sites that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,2,1,,,1,,organizational,1,0,The same 12 dentists reading the same radiographs without the software,67.43% tooth-level sensitivity,82.40% tooth-level sensitivity,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1740036818606034537.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:龋病口腔曲面体层图像辅助检测软件 (technical review report for the registration of dental caries panoramic radiograph detection-aid software), acceptance no. CQZ2400448, applicant 北京羽医甘蓝信息技术有限公司. 13 February 2025. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1740036818606034537.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:龋病口腔曲面体层图像辅助检测软件 (technical review report for the registration of dental caries panoramic radiograph detection-aid software), acceptance no. CQZ2400448, applicant 北京羽医甘蓝信息技术有限公司. 13 February 2025. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1740036818606034537.pdf""}]","{""kind"": ""Official documents""}" AIBD-0801,2025-01-24,source publication,day,tested,"FDA found a 3.5% rise in pathologists' sensitivity for prostate cancer with Galen Second Read, which flagged 202 of 347 cases called benign, of which experts confirmed 46","In FDA's review of Galen Second Read, pathologists' sensitivity rose from 90.5% to 93.9%. Their specificity fell from 91.1% to 87.9%. Of 347 cases called benign, it flagged 202. Experts confirmed 46.",*Galen Second Read* caught prostate cancers missed on a first read,pathologists' sensitivity rose from 90.5% to 93.9%,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Galen Second Read,3.1-US,Ibex Medical Analytics Ltd.,Israel,Ibex Medical Analytics (AIDER-2 study at three sites and AIDER-1 reader study with 12 pathologists at four sites),Israel,"Pathology sites in the United States, with one site outside the US in each study",United States,Healthcare,,Health,Images and video,Health workers,3,High income,physical_health,Medical diagnosis,,2,1,,,1,,organizational,2,0,The same 12 pathologists reading the same cases without the software,"Sensitivity 90.5%, specificity 91.1%","Sensitivity 93.9%, specificity 87.9%",measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/K241232.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. 510(k) substantial equivalence determination decision summary for Galen Second Read (K241232). Cleared 2025-01-24."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/K241232.pdf""}]","{""kind"": ""Official documents""}" AIBD-0655,2025-01-14,source publication,day,tested,NICE allows four systems in NHS urgent care after finding that AI fracture-detection software raised clinicians' sensitivity on X-rays,"NICE reviewed 16 studies of AI fracture software such as BoneView. It found that readers' sensitivity rose without lost specificity, from 61% to 74% in one study.",NICE supports AI help to find fractures on X-rays,from 61% to 74%,,found,Found by an assessment body,controlled comparison,very low,aggregate result,false,"AI fracture detection software: BoneView, RBfracture, Rayvolve and TechCare Alert",,"Gleamer (BoneView), Radiobotics (RBfracture), AZmed (Rayvolve) and Milvue (TechCare Alert)","Multiple: France, Denmark",Hospitals in the reader studies reviewed by NICE; NHS urgent care services in England during the evidence generation period,Multiple,"Multiple countries (studies); England, United Kingdom (guidance)",United Kingdom,Healthcare,,Health,Images and video,Health workers,3,High income,physical_health,Medical diagnosis,,,,3,,,,organizational,3,1,Emergency physicians reading mixed fracture types without BoneView,61% sensitivity,74% sensitivity,measured,persisted,2026-10-23,shown,https://www.nice.org.uk/guidance/htg739/chapter/3-Committee-discussion,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI) technologies to help detect fractures on X-rays in urgent care: early value assessment. HealthTech guidance HTG739, section 3: Committee discussion. 14 January 2025."", ""url"": ""https://www.nice.org.uk/guidance/htg739/chapter/3-Committee-discussion""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HTG739, section 1: Recommendations. 14 January 2025."", ""url"": ""https://www.nice.org.uk/guidance/htg739/chapter/1-Recommendations""}]","{""kind"": ""Official documents""}" @@ -199,11 +199,11 @@ AIBD-0926,2025-01-01,source publication,year,tested,"Avalia-t finds that the Wil AIBD-2310,2025-01-01,source publication,year,tested,SESCS's assessment: machine learning on FT-Raman spectra told platinum-resistant ovarian cancer from platinum-sensitive cancer with 95% to 100% accuracy in one small study,"In the one study that SESCS found, machine learning sorted FT-Raman spectra from 21 women by resistance to platinum with 95% to 100% accuracy. SESCS adds that the evidence is thin.",*FT-Raman* spectra with machine learning flag resistant ovarian cancer,95% to 100%,,found,Found by an assessment body,tested against a standard,very low,program,false,FT-Raman spectroscopy of blood serum or ovarian tissue with machine learning classifiers (a research method without a product name),,A research group in Poland (the authors of the study that SESCS reviewed),Poland,The research group in Poland that ran the study (no clinical use),Poland,Poland,Poland,Healthcare (ovarian cancer diagnosis),,Health,Prediction from data,Women and mothers; Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,personal,1,0,,,,,unknown,,shown,https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Espectroscopía FT-Raman con análisis de datos mediante aprendizaje automático para el diagnóstico de cáncer de ovario resistente al platino. Ficha de Evaluación de Tecnologías Sanitarias Nuevas y Emergentes. Ministerio de Sanidad and Servicio Canario de la Salud, 2025. SESCS is the evaluation service of the Canary Islands Health Service, and Spain's Ministry of Health funded the report. The report states that its conclusions are those of the service. The quotations are in Spanish."", ""url"": ""https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf""}, {""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Espectroscopía FT-Raman con análisis de datos mediante aprendizaje automático para el diagnóstico de cáncer de ovario resistente al platino. Ficha de Evaluación de Tecnologías Sanitarias Nuevas y Emergentes. Ministerio de Sanidad and Servicio Canario de la Salud, 2025. SESCS is the evaluation service of the Canary Islands Health Service, and Spain's Ministry of Health funded the report. The report states that its conclusions are those of the service. The quotations are in Spanish."", ""url"": ""https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf""}, {""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Espectroscopía FT-Raman con análisis de datos mediante aprendizaje automático para el diagnóstico de cáncer de ovario resistente al platino. Ficha de Evaluación de Tecnologías Sanitarias Nuevas y Emergentes. Ministerio de Sanidad and Servicio Canario de la Salud, 2025. SESCS is the evaluation service of the Canary Islands Health Service, and Spain's Ministry of Health funded the report. The report states that its conclusions are those of the service. The quotations are in Spanish."", ""url"": ""https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf""}, {""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Espectroscopía FT-Raman con análisis de datos mediante aprendizaje automático para el diagnóstico de cáncer de ovario resistente al platino. Ficha de Evaluación de Tecnologías Sanitarias Nuevas y Emergentes. Ministerio de Sanidad and Servicio Canario de la Salud, 2025. SESCS is the evaluation service of the Canary Islands Health Service, and Spain's Ministry of Health funded the report. The report states that its conclusions are those of the service. The quotations are in Spanish."", ""url"": ""https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf""}, {""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Espectroscopía FT-Raman con análisis de datos mediante aprendizaje automático para el diagnóstico de cáncer de ovario resistente al platino. Ficha de Evaluación de Tecnologías Sanitarias Nuevas y Emergentes. Ministerio de Sanidad and Servicio Canario de la Salud, 2025. SESCS is the evaluation service of the Canary Islands Health Service, and Spain's Ministry of Health funded the report. The report states that its conclusions are those of the service. The quotations are in Spanish."", ""url"": ""https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf""}, {""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Espectroscopía FT-Raman con análisis de datos mediante aprendizaje automático para el diagnóstico de cáncer de ovario resistente al platino. Ficha de Evaluación de Tecnologías Sanitarias Nuevas y Emergentes. Ministerio de Sanidad and Servicio Canario de la Salud, 2025. SESCS is the evaluation service of the Canary Islands Health Service, and Spain's Ministry of Health funded the report. The report states that its conclusions are those of the service. The quotations are in Spanish."", ""url"": ""https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf""}, {""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Espectroscopía FT-Raman con análisis de datos mediante aprendizaje automático para el diagnóstico de cáncer de ovario resistente al platino. Ficha de Evaluación de Tecnologías Sanitarias Nuevas y Emergentes. Ministerio de Sanidad and Servicio Canario de la Salud, 2025. SESCS is the evaluation service of the Canary Islands Health Service, and Spain's Ministry of Health funded the report. The report states that its conclusions are those of the service. The quotations are in Spanish."", ""url"": ""https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf""}, {""label"": ""Assessment report"", ""value"": ""Servicio de Evaluación del Servicio Canario de la Salud (SESCS). Espectroscopía FT-Raman con análisis de datos mediante aprendizaje automático para el diagnóstico de cáncer de ovario resistente al platino. Ficha de Evaluación de Tecnologías Sanitarias Nuevas y Emergentes. Ministerio de Sanidad and Servicio Canario de la Salud, 2025. SESCS is the evaluation service of the Canary Islands Health Service, and Spain's Ministry of Health funded the report. The report states that its conclusions are those of the service. The quotations are in Spanish."", ""url"": ""https://sescs.es/wp-content/uploads/2025/05/TE_2023SSTE40_2024_SESCS_FT-raman_cancer_ovario_DEF_NIPO.pdf""}]","{""kind"": ""Official documents""}" AIBD-1060,2024-12-16,source publication,day,tested,China's CMDE finds that endoscopists using HW-ES missed fewer colon polyps in the maker's randomized trial,"In the maker's randomized trial that China's CMDE reviewed, endoscopists using HW-ES missed 12.41% of polyps, against 50.80% without it.",*HW-ES* helped endoscopists miss fewer colon polyps in a trial,"missed 12.41% of polyps, against 50.80% without it",,found,Found by a regulator,randomized trial,moderate,aggregate result,false,HW-ES (肠息肉电子下消化道内窥镜图像辅助检测软件),Release V1,"Changsha Huiwei Intelligent Medical Technology Co., Ltd. (长沙慧维智能医疗科技有限公司)",China,Changsha Huiwei Intelligent Medical Technology (prospective randomized trial with 354 patients at three sites),China,"Three trial sites that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,3,1,,,1,,organizational,1,0,Endoscopists examining without the software in the same randomized trial,50.80% lesion-level polyp miss rate,12.41% lesion-level polyp miss rate,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1742969333809034777.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:肠息肉电子下消化道内窥镜图像辅助检测软件 (technical review report for the registration of colon polyp detection-aid software for electronic lower gastrointestinal endoscope images), acceptance no. CQZ2302046, applicant 长沙慧维智能医疗科技有限公司. 16 December 2024. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1742969333809034777.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:肠息肉电子下消化道内窥镜图像辅助检测软件 (technical review report for the registration of colon polyp detection-aid software for electronic lower gastrointestinal endoscope images), acceptance no. CQZ2302046, applicant 长沙慧维智能医疗科技有限公司. 16 December 2024. Benefit-risk determination and overall opinion, CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1742969333809034777.pdf""}]","{""kind"": ""Official documents""}" AIBD-0744,2024-12-01,source publication,month,mixed,"CADTH, now Canada's Drug Agency, finds that AI-assisted colonoscopy, which is likely cost-effective, may detect more adenomas","CADTH reviewed five studies, mostly systematic reviews. It found that AI-assisted colonoscopy may detect more adenomas. It may also miss fewer. CADTH found that it is likely cost-effective.",CADTH finds that *AI-assisted colonoscopy* may detect more adenomas,,,found,Found by an assessment body,controlled comparison,very low,aggregate result,false,"AI-assisted (computer-aided detection) colonoscopy systems, including Deep-GI and CAD EYE in the included trial",,"Several makers, including Fujifilm (CAD EYE)",Multiple countries (makers not named),Endoscopy services in the studies that CADTH reviewed,Multiple,"Multiple countries (studies), with guidance for Canada",Canada,Healthcare,,Health,Images and video,Health workers,3,High income,physical_health,Medical diagnosis,,,,,2,,,personal,2,0,Conventional colonoscopy,,,measured,challenged,2025-08-12,shown,https://healthtechnology.wales/wp-content/uploads/EAR055-AIAE-WEB.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Assessment report"", ""value"": ""Canadian Agency for Drugs and Technologies in Health. Artificial Intelligence–Assisted Colonoscopy for Detecting Polyps, Adenomas, Precancerous Lesions, and Colorectal Cancer. CADTH Health Technology Review, Report No. RC1559, December 2024. NCBI Bookshelf NBK611312, PMID 39869728 (key messages)."", ""url"": ""https://europepmc.org/abstract/MED/39869728""}, {""label"": ""Assessment report"", ""value"": ""Canadian Agency for Drugs and Technologies in Health. CADTH Health Technology Review RC1559, December 2024, key messages (economic evidence)."", ""url"": ""https://europepmc.org/abstract/MED/39869728""}, {""label"": ""Assessment report"", ""value"": ""Behandlingsrådet (Danish Health Technology Council). Kunstig intelligens som beslutningsstøtte ved koloskopiske undersøgelser til diagnosticering af neoplastisk sygdom. Recommendation in force from 1 February 2023. Archived copy of 15 September 2025 of the council's page. The quotations are in Danish."", ""url"": ""https://web.archive.org/web/20250915185757id_/https://behandlingsraadet.dk/find-evalueringer-analyser/analyser/2021/kunstig-intelligens-som-beslutningsstotte-ved-koloskopiske-undersogelser-til-diagnosticering-af-neoplastisk-sygdom""}, {""label"": ""Assessment report"", ""value"": ""Behandlingsrådet (Danish Health Technology Council). Kunstig intelligens som beslutningsstøtte ved koloskopiske undersøgelser til diagnosticering af neoplastisk sygdom. Recommendation in force from 1 February 2023. Archived copy of 15 September 2025 of the council's page. The quotations are in Danish."", ""url"": ""https://web.archive.org/web/20250915185757id_/https://behandlingsraadet.dk/find-evalueringer-analyser/analyser/2021/kunstig-intelligens-som-beslutningsstotte-ved-koloskopiske-undersogelser-til-diagnosticering-af-neoplastisk-sygdom""}, {""label"": ""Follow-up"", ""value"": ""Polish researchers reported in August 2025 that endoscopists' unassisted adenoma detection fell from 28.4% to 22.4% after three months of routine AI use, raising a deskilling concern. The study was observational."", ""url"": ""https://wrap.warwick.ac.uk/id/eprint/191005""}]","{""kind"": ""Challenges""}" -AIBD-1212,2024-11-26,source publication,day,intended,Japan's Kanto District Transport Bureau approves an Ibaraki Kotsu bus on the Hitachi BRT as a Level 4 automated vehicle that needs no driver,On 26 November 2024 a Japanese transport bureau confirmed that a Hitachi BRT bus's automated operation device met the safety standards. It approved the bus as a Level 4 vehicle on about 6.1 km.,*Automated operation device* lets a Hitachi BRT bus run with no driver,,,found,Found by a regulator,,not rated,benefit created,false,Automated operation device (自動運行装置) on an Isuzu Erga Mio bus for the Hitachi BRT,,"Advanced Smart Mobility Co., Ltd. (先進モビリティ株式会社), the applicant",Japan,"Ibaraki Kotsu Co., Ltd. (茨城交通株式会社), the operator",Japan,"Hitachi, Ibaraki Prefecture, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,physical_health; operational_improvement,,,,,,,,,societal,4,0,,,,,persisted,2025-01-24,not shown,https://www.soumu.go.jp/main_content/001034209.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 茨城県初!自動運転車(レベル4)の認可を行いました (Ibaraki's first approval of a Level 4 automated vehicle). Press release, 26 November 2024."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000336532.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 茨城県初!自動運転車(レベル4)の認可を行いました (Ibaraki's first approval of a Level 4 automated vehicle). Press release, 26 November 2024."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000336532.pdf""}]","{""kind"": ""Official documents""}" +AIBD-1212,2024-11-26,source publication,day,intended,Japan's Kanto District Transport Bureau approves an Ibaraki Kotsu bus on the Hitachi BRT as a Level 4 automated vehicle that needs no driver,On 26 November 2024 a Japanese transport bureau confirmed that a Hitachi BRT bus's automated operation device met the safety standards. It approved the bus as a Level 4 vehicle on about 6.1 km.,*Automated operation device* lets a Hitachi BRT bus run with no driver,,,found,Found by a regulator,,very low,benefit created,false,Automated operation device (自動運行装置) on an Isuzu Erga Mio bus for the Hitachi BRT,,"Advanced Smart Mobility Co., Ltd. (先進モビリティ株式会社), the applicant",Japan,"Ibaraki Kotsu Co., Ltd. (茨城交通株式会社), the operator",Japan,"Hitachi, Ibaraki Prefecture, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,physical_health; operational_improvement,,,,,,,,,societal,4,0,,,,,persisted,2025-01-24,not shown,https://www.soumu.go.jp/main_content/001034209.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 茨城県初!自動運転車(レベル4)の認可を行いました (Ibaraki's first approval of a Level 4 automated vehicle). Press release, 26 November 2024."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000336532.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 茨城県初!自動運転車(レベル4)の認可を行いました (Ibaraki's first approval of a Level 4 automated vehicle). Press release, 26 November 2024."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000336532.pdf""}]","{""kind"": ""Official documents""}" AIBD-0745,2024-11-24,event,day,reported,"Medicare contractor Palmetto GBA covers AI coronary plaque analysis for stable chest pain, citing better diagnosis","Palmetto GBA found AI coronary plaque analysis reasonable and necessary for chest pain without known coronary disease. In a cited study, physicians changed plans for 57.1% of patients.",Medicare contractor covers *AI-QCT* plaque analysis for chest pain,changed management plans for 57.1% of 750 patients,,found,Found by an assessment body,controlled comparison,low,aggregate result,false,"AI-enabled CT-based quantitative coronary topography and coronary plaque analysis (AI-QCT/AI-CPA), chiefly Cleerly's platform, with HeartFlow plaque analysis also in the evidence",,"Cleerly, Inc. and HeartFlow, Inc.",United States,Medicare providers performing coronary CT angiography in Palmetto GBA's jurisdictions,United States,"Alabama, Georgia, Tennessee, South Carolina, North Carolina, Virginia and West Virginia, United States",United States,Healthcare (cardiology),,Health,Images and video,Patients; Health workers,3,High income,physical_health,Medical diagnosis,,3,1,,3,1,,organizational; personal,9,0,Physicians' management plans for the same patients based on CT angiography alone (CERTAIN study),Management plans made from the CT angiography results,a change in management for 57.1% (428/750) of patients,measured,persisted,2026-10-15,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf20/K202280.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Palmetto GBA. Local Coverage Determination L39851: Artificial Intelligence Enabled CT Based Quantitative Coronary Topography (AI-QCT)/Coronary Plaque Analysis (AI-CPA). Original effective date 24 November 2024, revision effective 15 October 2026. CMS Medicare Coverage Database, Analysis of Evidence."", ""url"": ""https://www.cms.gov/medicare-coverage-database/view/lcd.aspx?lcdid=39851&ver=18""}, {""label"": ""Assessment report"", ""value"": ""Wellpoint Federal (formerly National Government Services), Medicare Administrative Contractor for CMS. LCD L39863: Artificial Intelligence Enabled CT Based Quantitative Coronary Topography (AI-QCT)/Coronary Plaque Analysis (AI-CPA). Analysis of Evidence (Rationale for Determination)"", ""url"": ""https://www.cms.gov/medicare-coverage-database/view/lcd.aspx?lcdid=39863&ver=10""}, {""label"": ""Assessment report"", ""value"": ""UnitedHealthcare (commercial and exchange plans), Cardiovascular and Radiology Imaging Guidelines (2026). Cardiovascular and Radiology Imaging Guidelines V6.0.2026, Cardiac Imaging: Coronary CTA plaque quantification (CPT 75577) (effective 1 October 2026), Cardiac Imaging Guidelines, 'Coronary CTA plaque quantification CPT 75577' (indications and non-indications); references 50-51."", ""url"": ""https://www.uhcprovider.com/content/dam/provider/docs/public/prior-auth/r-c/COMM-Exchange-Rad-Card-Guidelines-10-2026.pdf""}, {""label"": ""Assessment report"", ""value"": ""Noridian Healthcare Solutions (Medicare Administrative Contractor, Jurisdictions E and F) (2024). LCD L39881: Artificial Intelligence Enabled CT Based Quantitative Coronary Topography (AI-QCT)/Coronary Plaque Analysis (AI-CPA), Summary of Evidence and Analysis of Evidence (Rationale for Determination)."", ""url"": ""https://www.cms.gov/medicare-coverage-database/view/lcd.aspx?lcdid=39881""}, {""label"": ""Assessment report"", ""value"": ""Wisconsin Physicians Service Insurance Corporation (WPS, Medicare Administrative Contractor, Jurisdictions 5 and 8) (2024). LCD L39913: Artificial Intelligence Enabled CT Based Quantitative Coronary Topography (AI-QCT)/Coronary Plaque Analysis (AI-CPA), Summary of Evidence and Analysis of Evidence (Rationale for Determination)."", ""url"": ""https://www.cms.gov/medicare-coverage-database/view/lcd.aspx?lcdid=39913""}, {""label"": ""Assessment report"", ""value"": ""CGS Administrators (Medicare Administrative Contractor, Jurisdiction 15) (2024). LCD L39840: Artificial Intelligence Enabled CT Based Quantitative Coronary Topography (AI-QCT)/Coronary Plaque Analysis (AI-CPA), Summary of Evidence and Analysis of Evidence (Rationale for Determination)."", ""url"": ""https://www.cms.gov/medicare-coverage-database/view/lcd.aspx?lcdid=39840""}, {""label"": ""Assessment report"", ""value"": ""UnitedHealthcare Community Plan (Medicaid managed care), Cardiovascular and Radiology Imaging Guidelines (2026). UnitedHealthcare Community Plan Cardiovascular and Radiology Imaging Guidelines V6.0.2026, Coronary CTA plaque quantification (effective 1 October 2026), Cardiac Imaging Guidelines, 'Coronary CTA plaque quantification' (indications and non-indications)."", ""url"": ""https://www.uhcprovider.com/content/dam/provider/docs/public/prior-auth/r-c/UHCCP-Rad-Card-Guidelines-Oct-2026.pdf""}, {""label"": ""Assessment report"", ""value"": ""Oregon Health Evidence Review Commission (HERC), Oregon Health Authority, Value-based Benefits Subcommittee (VbBS) (2025). HERC/VbBS meeting materials, 9 October 2025: 2026 CPT Code Review, AI Coronary Atherosclerotic Plaque Assessment (CPT 75577, replacing 0623T-0626T), 2026 CPT Code Review: AI Coronary Atherosclerotic Plaque Assessment, evidence and HERC staff summary."", ""url"": ""https://www.oregon.gov/oha/HPA/DSI-HERC/MeetingDocuments/HERC-VBBS-Materials-10-9-2025.pdf""}, {""label"": ""Assessment report"", ""value"": ""Kaiser Foundation Health Plan of Washington (Kaiser Permanente), Medical Policy Committee, Clinical Review Criteria (2026). Clinical Review Criteria: Cardiac CT, Coronary CT Angiography, Calcium Scoring and CT Fractional Flow Reserve, including AI-QCT/AI-CPA (reviewed 5 May 2026, revised 10 July 2026), Criteria, 'For Non-Medicare Members', row for Artificial Intelligence Enabled CT Based Quantitative Coronary Topography (AI-QCT)/Coronary Plaque Analysis (AI-CPA) (CPT 75577)."", ""url"": ""https://wa-provider.kaiserpermanente.org/static/pdf/hosting/clinical/criteria/pdf/ct_angiography_diagnostic.pdf""}]","{""kind"": ""Official documents""}" AIBD-1084,2024-11-04,source publication,day,tested,China's CMDE finds that uAI-AorticDissection flagged aortic dissection on CT angiography with 99.46% sensitivity in the maker's trial,"In the maker's retrospective trial that China's CMDE reviewed, uAI-AorticDissection flagged aortic dissection with 99.46% sensitivity and 98.45% specificity against an expert panel.",*uAI-AorticDissection* flagged aortic dissections in China's review,99.46% sensitivity and 98.45% specificity,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,uAI-AorticDissection (主动脉夹层CT血管造影图像辅助分诊软件),Release 1.0 (full version 1.0.0.0.18),"Shanghai United Imaging Intelligence Co., Ltd. (上海联影智能医疗科技有限公司)",China,Shanghai United Imaging Intelligence (retrospective trial with 380 subjects at three institutions),China,Three trial institutions in China that the report does not name,China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,"Targets set in the trial protocol, with an expert panel's reading as the reference",Protocol targets that the report does not state,99.46% sensitivity and 98.45% specificity,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1732498238794063154.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:主动脉夹层CT血管造影图像辅助分诊软件 (technical review report for the registration of aortic dissection CT angiography image triage-aid software), acceptance no. CQZ2400640, applicant 上海联影智能医疗科技有限公司. 4 November 2024. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1732498238794063154.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医��器械产品注册技术审评报告:主动脉夹层CT血管造影图像辅助分诊软件 (technical review report for the registration of aortic dissection CT angiography image triage-aid software), acceptance no. CQZ2400640, applicant 上海联影智能医疗科技有限公司. 4 November 2024. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1732498238794063154.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0951,2024-10-31,source publication,day,intended,Japan's Hokuriku-Shinetsu District Transport Bureau approves TIER IV's automated minibus in Shiojiri as a Level 4 vehicle that needs no driver,On 31 October 2024 a Japanese transport bureau confirmed that TIER IV's automated operation device met the safety standards. It approved a Shiojiri minibus as a Level 4 vehicle with no driver.,Japan approves *AI Pilot* to drive a Shiojiri minibus with no driver,,,found,Found by a regulator,,not rated,benefit created,false,Automated operation device (TIER IV's AI Pilot) on the TIER IV Minibus,,"TIER IV, Inc.",Japan,"Shiojiri City Promotion Corporation (一般財団法人塩尻市振興公社), the operator",Japan,"Shiojiri, Nagano Prefecture, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,operational_improvement; physical_health,Enhanced productivity,,,,,,,,societal,4,0,,,,,persisted,2026-11-02,not shown,https://web.archive.org/web/20260209202144id_/https://technode.global/prnasia/tier-ivs-ai-pilot-system-certified-for-level-4-autonomous-bus-service-in-nagano-prefecture/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Hokuriku-Shinetsu District Transport Bureau, Automotive Technology and Safety Department. 北陸信越運輸局管内初!!自動運転車(レベル4)の認可について (First in the bureau's area: approval of a Level 4 automated vehicle). Press release, 31 October 2024, paragraphs 1 to 3 and table."", ""url"": ""https://wwwtb.mlit.go.jp/hokushin/content/000335753.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0992,2024-10-21,source publication,day,intended,"The Delaware Court of Chancery let defendants review 125,000 documents with technology assisted review over a demand for manual review","In October 2024 the Delaware Court of Chancery let defendants review 125,000 documents with technology assisted review over the plaintiff's demand for manual review. The review had not run.",Delaware court lets defendants use *TAR* over a demand for manual review,"125,000 documents",,found,Found by a court,,not rated,benefit created,false,"Technology assisted review (TAR), or predictive coding",,Not named in the opinion,,The former directors and officers of Graf Industrial Corp. (defendants),United States,"Wilmington, Delaware, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,Manual review of all the documents captured by the plaintiff's search terms,,,assumed,unknown,,shown,https://courts.delaware.gov/Opinions/Download.aspx?id=371000,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Court of Chancery of the State of Delaware (Vice Chancellor Lori W. Will). Berger v. Graf Acquisition, LLC, et al., C.A. No. 2023-0873-LWW, letter opinion of 21 October 2024 on the plaintiff's motion to compel, pp. 4-9 and 12."", ""url"": ""https://courts.delaware.gov/Opinions/Download.aspx?id=371000""}]","{""kind"": ""Official documents""}" +AIBD-0951,2024-10-31,source publication,day,intended,Japan's Hokuriku-Shinetsu District Transport Bureau approves TIER IV's automated minibus in Shiojiri as a Level 4 vehicle that needs no driver,On 31 October 2024 a Japanese transport bureau confirmed that TIER IV's automated operation device met the safety standards. It approved a Shiojiri minibus as a Level 4 vehicle with no driver.,Japan approves *AI Pilot* to drive a Shiojiri minibus with no driver,,,found,Found by a regulator,,very low,benefit created,false,Automated operation device (TIER IV's AI Pilot) on the TIER IV Minibus,,"TIER IV, Inc.",Japan,"Shiojiri City Promotion Corporation (一般財団法人塩尻市振興公社), the operator",Japan,"Shiojiri, Nagano Prefecture, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,operational_improvement; physical_health,Enhanced productivity,,,,,,,,societal,4,0,,,,,persisted,2026-11-02,not shown,https://web.archive.org/web/20260209202144id_/https://technode.global/prnasia/tier-ivs-ai-pilot-system-certified-for-level-4-autonomous-bus-service-in-nagano-prefecture/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Hokuriku-Shinetsu District Transport Bureau, Automotive Technology and Safety Department. 北陸信越運輸局管内初!!自動運転車(レベル4)の認可について (First in the bureau's area: approval of a Level 4 automated vehicle). Press release, 31 October 2024, paragraphs 1 to 3 and table."", ""url"": ""https://wwwtb.mlit.go.jp/hokushin/content/000335753.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0992,2024-10-21,source publication,day,intended,"The Delaware Court of Chancery let defendants review 125,000 documents with technology assisted review over a demand for manual review","In October 2024 the Delaware Court of Chancery let defendants review 125,000 documents with technology assisted review over the plaintiff's demand for manual review. The review had not run.",Delaware court lets defendants use *TAR* over a demand for manual review,"125,000 documents",,found,Found by a court,,very low,benefit created,false,"Technology assisted review (TAR), or predictive coding",,Not named in the opinion,,The former directors and officers of Graf Industrial Corp. (defendants),United States,"Wilmington, Delaware, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,Manual review of all the documents captured by the plaintiff's search terms,,,assumed,unknown,,shown,https://courts.delaware.gov/Opinions/Download.aspx?id=371000,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Court of Chancery of the State of Delaware (Vice Chancellor Lori W. Will). Berger v. Graf Acquisition, LLC, et al., C.A. No. 2023-0873-LWW, letter opinion of 21 October 2024 on the plaintiff's motion to compel, pp. 4-9 and 12."", ""url"": ""https://courts.delaware.gov/Opinions/Download.aspx?id=371000""}]","{""kind"": ""Official documents""}" AIBD-1063,2024-09-29,source publication,day,tested,China's CMDE finds that doctors using Dr.Wise AI Lung Cancer judged the malignancy of lung nodules more accurately in the maker's trial,"In the maker's trial that China's CMDE reviewed, doctors' area under the ROC curve for suspected malignant lung nodules rose from 0.7325 to 0.8397 with Dr.Wise AI Lung Cancer.",*Dr.Wise AI Lung Cancer* helped doctors rate risk of lung nodules on CT,area under the ROC curve for suspected malignant lung nodules rose from 0.7325 to 0.8397,,found,Found by a regulator,controlled comparison,very low,aggregate result,false,Dr.Wise AI Lung Cancer (肺结节CT图像辅助诊断软件),Release 1 (full version 1.1.0.1),Beijing Deepwise (北京深睿博联科技有限责任公司),China,Beijing Deepwise (multi-reader trial with 400 subjects at three sites),China,"Three trial sites that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,,1,,,1,,organizational,1,0,"The same doctors reading the same CT cases without the software, with imaging, thoracic surgery and pathology experts as the reference",Subject-level area under the ROC curve 0.7325±0.0234,Subject-level area under the ROC curve 0.8397±0.0165,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1729664104391037246.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:肺结节CT图像辅助诊断软件 (technical review report for the registration of lung nodule diagnosis-aid software for CT images), acceptance no. CQZ2302309, applicant 北京深睿博联科技有限责任公司. 29 September 2024. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1729664104391037246.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:肺结节CT图像辅助诊断软件 (technical review report for the registration of lung nodule diagnosis-aid software for CT images), acceptance no. CQZ2302309, applicant 北京深睿博联科技有限责任公司. 29 September 2024. Benefit-risk determination and overall opinion, CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1729664104391037246.pdf""}]","{""kind"": ""Official documents""}" AIBD-1018,2024-09-23,source publication,day,tested,"CATARC found in C-ICAP tests that ADS Pro V2.0 on the AITO M9 scored 100.0% for emergency avoidance, with 95.8% for navigation-guided driving","In CATARC's C-ICAP tests, published on 23 September 2024, ADS Pro V2.0 on an AITO M9 scored 100.0% for emergency avoidance. It scored 95.8% for navigation-guided driving assistance.",*ADS Pro V2.0* scores 100.0% for emergency avoidance on an AITO M9,100.0% for emergency avoidance,,found,Found by an assessment body,no comparison,very low,aggregate result,false,"ADS Pro V2.0 with the ADS high-level function pack (ADS高阶功能包), as fitted to the AITO M9 (问界M9) Ultra BEV",ADS Pro V2.0(2024.24.1eAcK8T3K0Ye) (software version tested),"Huawei (ADS), fitted by Seres on the AITO M9",China,"Seres (重庆赛力斯汽车有限公司), factory-fitted on the AITO M9",China,China,China,Road vehicle crash avoidance / assisted driving,,Safety and rescue,Machines that act,Drivers and travelers,11,Upper middle income,physical_health,Hazard reduction,,,,,,,,personal,1,0,,,,,unknown,,shown,https://carnewschina.com/2024/04/24/huawei-qiankun-ads-3-0-announced-brings-intelligent-driving-upgrade/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""China Automotive Technology and Research Center (CATARC), Automotive Test & Assessment Management Center. C-ICAP result for the 赛力斯问界M9 Ultra纯电版 (AITO M9 Ultra BEV), protocol version 1.1, test year 2024. Result page published 23 September 2024, with CATARC's rating report linked from it."", ""url"": ""https://www.c-ncap.org.cn/evaluation/e45fb38aacbd2f6e4afd9f24728b0a14""}]","{""kind"": ""Official documents""}" AIBD-1019,2024-09-23,source publication,day,tested,CATARC found in C-ICAP tests that XPeng's assisted driving on the X9 scored 100.0% for emergency avoidance but had no minimal risk maneuver,"In CATARC's C-ICAP tests, published on 23 September 2024, XPeng's assisted driving on an X9 scored 100.0% for emergency avoidance. It had no minimal risk maneuver.",*XOS 5.0.5* scores 100.0% for emergency avoidance on an XPeng X9,100.0% for emergency avoidance,,found,Found by an assessment body,no comparison,very low,aggregate result,false,"XPeng assisted driving in XOS 5.0.5, as fitted to the XPeng X9 (小鹏X9) 702 Max",XOS 5.0.5 (software version tested),XPeng (小鹏),China,"XPeng (肇庆小鹏新能源投资有限公司), factory-fitted on the X9",China,China,China,Road vehicle crash avoidance / assisted driving,,Safety and rescue,Machines that act,Drivers and travelers,11,Upper middle income,physical_health,Hazard reduction,,,,,,,,personal,1,0,,,,,unknown,,shown,https://cnevpost.com/2022/10/24/xpeng-tech-day-2022/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""China Automotive Technology and Research Center (CATARC), Automotive Test & Assessment Management Center. C-ICAP result for the 小鹏X9 702 超长续航Max (XPeng X9 702 Max), protocol version 1.1, test year 2024. Result page published 23 September 2024, with CATARC's rating report linked from it."", ""url"": ""https://www.c-ncap.org.cn/evaluation/8fef19bc48306f3d021dd3c7479b9bbb""}]","{""kind"": ""Official documents""}" @@ -217,7 +217,7 @@ AIBD-1062,2024-07-18,source publication,day,tested,China's CMDE finds that docto AIBD-1023,2024-06-19,source publication,day,tested,CATARC found in C-ICAP tests that the Jiyue 01's assisted driving scored 100.0% for emergency avoidance but collided when a car ahead moved aside,"In CATARC's C-ICAP tests, published on 19 June 2024, the Jiyue 01 Max scored 100.0% for emergency avoidance. It collided when a car ahead moved aside to reveal a stopped car.",*Jiyue 01 assisted driving* scores 100.0% for emergency avoidance,100.0% for emergency avoidance,,found,Found by an assessment body,no comparison,very low,aggregate result,false,"Jiyue 01 assisted driving, software V1.3.1, as fitted to the Jiyue 01 (极越01) Max (the page does not name the system)",V1.3.1 (software version tested),"Jiyue (上海集度汽车), with Baidu Apollo (not named on the page)",China,"Jiyue (上海集度汽车有限公司), factory-fitted on the Jiyue 01",China,China,China,Road vehicle crash avoidance / assisted driving,,Safety and rescue,Machines that act,Drivers and travelers,11,Upper middle income,physical_health,Hazard reduction,,,,,,,,personal,1,0,,,,,withdrawn,2024-12-11,shown,https://autonews.gasgoo.com/articles/news/ji-yue-introduces-pure-vision-based-intelligent-driving-solution-70028988,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Assessment report"", ""value"": ""China Automotive Technology and Research Center (CATARC), Automotive Test & Assessment Management Center. C-ICAP result for the 极越01 Max (Jiyue 01 Max), protocol version 1.1, test year 2024. Result page published 19 June 2024, with CATARC's rating report linked from it."", ""url"": ""https://www.c-ncap.org.cn/evaluation/b71f517eb3fede6f2bfbcf03f466a43d""}, {""label"": ""Follow-up"", ""value"": ""On 11 December 2024 Jiyue Auto, the carmaker backed by Baidu and Geely that made the Jiyue 01, told employees that it was shutting down entirely."", ""url"": ""https://en.jiemian.com/article/12189660.html""}]","{""kind"": ""Withdrawals""}" AIBD-1094,2024-06-11,source publication,day,tested,China's CMDE finds that the MCS fundus software met its targets for accuracy on 11 eye diseases in the maker's trial,"In the maker's trial that China's CMDE reviewed, the MCS fundus software met its targets for 11 eye diseases, with 91.18% sensitivity and 91.56% specificity for diabetic retinopathy.",*MCS* flagged 11 eye diseases on fundus photos in China's review,91.18% sensitivity and 91.56% specificity for diabetic retinopathy,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,MCS / MCSLocal (眼底病变眼底图像辅助诊断软件),Release V2 (MCS cloud server version and MCSLocal local server version),"Beijing Zhiyuan Huitu Technology Co., Ltd. (北京致远慧图科技有限公司)",China,"Beijing Zhiyuan Huitu Technology (registration trial with 2,778 fundus photographs at four institutions)",China,"Four trial institutions that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,"Targets set for each disease from a pilot trial and the literature, with central reading as the gold standard",Target of 82.0% sensitivity for diabetic retinopathy,"91.18% [86.48%,94.35%] sensitivity for diabetic retinopathy",measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1742968101004034249.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:眼底病变眼底图像辅助诊断软件 (technical review report for the registration of fundus disease fundus image diagnosis-aid software), acceptance no. CQZ2200072, applicant 北京致远慧图科技有限公司. 11 June 2024. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1742968101004034249.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:眼底病变眼底图像辅助诊断软件 (technical review report for the registration of fundus disease fundus image diagnosis-aid software), acceptance no. CQZ2200072, applicant 北京致远慧图科技有限公司. 11 June 2024. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1742968101004034249.pdf""}]","{""kind"": ""Official documents""}" AIBD-0827,2024-06-01,event,month,tested,EPA approves SENSIA's RedLook System for compliance after finding that it meets the 1 kg per hour threshold for screening methane leaks,EPA found in June 2026 that SENSIA's RedLook System meets the 1 kg/hr threshold for periodic screening for methane. It uses machine learning to analyze camera video.,*RedLook* software meets EPA's threshold of 1 kg per hour for methane,meets the 1 kg/hr threshold,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,"RedLook System, with the Caroline FY and Mileva cameras for optical gas imaging",,"SENSIA (Leganés, Madrid)",Spain,Oil and gas owners and operators at affected facilities (approved use),United States,United States,United States,Climate change mitigation / methane emissions monitoring,,Environment,Images and video,,13,High income,environmental_impact,Environmental impacts,,,,,,,,societal,3,0,The 1 kg/hr threshold for periodic screening,1 kg/hr,"90% probability of detection at 0.84 kg/hr, without wind correction",measured,unknown,,shown,https://www.epa.gov/system/files/documents/2026-08/sensia-mileva-approval-letter.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Environmental Protection Agency, Office of Air and Radiation. Approval letter for the SENSIA RedLook System Caroline FY as Methane Alternative Test Method MATM-017. 12 June 2026."", ""url"": ""https://www.epa.gov/system/files/documents/2026-08/sensia-caroline-approval-letter.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Environmental Protection Agency, Office of Air and Radiation. Approval letter for the SENSIA RedLook System Mileva as Methane Alternative Test Method MATM-018. 12 June 2026."", ""url"": ""https://www.epa.gov/system/files/documents/2026-08/sensia-mileva-approval-letter.pdf""}]","{""kind"": ""Official documents""}" -AIBD-1015,2024-05-31,source publication,day,intended,Japan's Hokkaido District Transport Bureau approves a NAVYA ARMA shuttle in Kamishihoro as a Level 4 automated vehicle that needs no driver,On 31 May 2024 a Japanese transport bureau approved a NAVYA ARMA shuttle in Kamishihoro for Level 4. It confirmed that the shuttle's automated operation device met the safety standards.,NAVYA's *ARMA* gets approval to run with no driver in Kamishihoro,,,found,Found by a regulator,,not rated,benefit created,false,Automated operation device on a NAVYA ARMA shuttle,,NAVYA (maker of the ARMA shuttle),France,"BOLDLY Inc. (BOLDLY株式会社), the applicant and operator, in the town of Kamishihoro",Japan,"Kamishihoro, Hokkaido, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,physical_health; operational_improvement,,,,,,,,,societal,2,0,,,,,persisted,2024-10-25,shown,https://www.anyauto.com.au/?p=16136,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Hokkaido District Transport Bureau, Automotive Technology and Safety Department. 北海道初! 自動運転車(レベル4)の認可について ~バス運転手不足の解決に一歩前進~ (First in Hokkaido: approval of a Level 4 automated vehicle, a step toward solving the bus driver shortage). Press release, 31 May 2024, paragraphs 1 and 2, table and attachment."", ""url"": ""https://wwwtb.mlit.go.jp/hokkaido/content/000325959.pdf""}]","{""kind"": ""Official documents""}" +AIBD-1015,2024-05-31,source publication,day,intended,Japan's Hokkaido District Transport Bureau approves a NAVYA ARMA shuttle in Kamishihoro as a Level 4 automated vehicle that needs no driver,On 31 May 2024 a Japanese transport bureau approved a NAVYA ARMA shuttle in Kamishihoro for Level 4. It confirmed that the shuttle's automated operation device met the safety standards.,NAVYA's *ARMA* gets approval to run with no driver in Kamishihoro,,,found,Found by a regulator,,very low,benefit created,false,Automated operation device on a NAVYA ARMA shuttle,,NAVYA (maker of the ARMA shuttle),France,"BOLDLY Inc. (BOLDLY株式会社), the applicant and operator, in the town of Kamishihoro",Japan,"Kamishihoro, Hokkaido, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,physical_health; operational_improvement,,,,,,,,,societal,2,0,,,,,persisted,2024-10-25,shown,https://www.anyauto.com.au/?p=16136,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Hokkaido District Transport Bureau, Automotive Technology and Safety Department. 北海道初! 自動運転車(レベル4)の認可について ~バス運転手不足の解決に一歩前進~ (First in Hokkaido: approval of a Level 4 automated vehicle, a step toward solving the bus driver shortage). Press release, 31 May 2024, paragraphs 1 and 2, table and attachment."", ""url"": ""https://wwwtb.mlit.go.jp/hokkaido/content/000325959.pdf""}]","{""kind"": ""Official documents""}" AIBD-0772,2024-05-17,source publication,day,tested,"In FDA's De Novo review of DentalMonitoring, its AI read orthodontic photos taken by phone in agreement with panels of orthodontists","In studies that FDA reviewed, DentalMonitoring's analysis of patients' phone scans matched orthodontist panels. For black triangles, for example, sensitivity was 81.0%. Specificity was 98.4%.",*DentalMonitoring* reads phone photos of braces for orthodontists,sensitivity was 81.0% and specificity 98.4%,,found,Found by a regulator,no comparison,very low,aggregate result,false,DentalMonitoring,,Dental Monitoring SAS,France,Dental Monitoring (retrospective and prospective studies with US orthodontists),France,United States (retrospective review with 15 US orthodontists and prospective studies at US sites),United States,Healthcare,,Health,Images and video,Health workers; Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,,,,,persisted,2025-06-20,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230035.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for DentalMonitoring (DEN230035): decision summary. Granted 2024-05-17."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230035.pdf""}]","{""kind"": ""Official documents""}" AIBD-0768,2024-05-15,source publication,day,tested,"In FDA's De Novo review of the Notal Vision Home OCT, patients' home retinal scans agreed with clinic scans on fluid in most pairs","In the five-week study that FDA reviewed, patients' home OCT scans agreed with clinic scans on retinal fluid, with a PPA of 0.864. The NPA was 0.849. No gain in vision was shown.",*Notal Vision Home OCT* lets patients scan their own retinas at home,a positive percent agreement of 0.864,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Notal Vision Home Optical Coherence Tomography (OCT) System with the Notal OCT Analyzer (NOA),,"Notal Vision, Inc.",United States,"Notal Vision (001 Study at seven US sites, with patients imaging themselves at home, and 006 Study at six US sites)",United States,Patients' homes and seven US clinical sites (001 Study) and six US sites (006 Study),United States,Healthcare,,Health,Images and video,Patients,3,High income,physical_health,Medical access,,3,2,,,5,,personal,2,0,Clinic OCT scans graded by the reading center within 24 hours,Fluid status on clinic OCT,PPA 0.864 and NPA 0.849 for home scans,measured,persisted,2026-01-01,shown,https://www.ophthalmologytimes.com/view/notal-vision-shares-data-on-novel-home-oct-device,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for Notal Vision Home Optical Coherence Tomography (OCT) System (DEN230043): decision summary. Granted 2024-05-15."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230043.pdf""}]","{""kind"": ""Official documents""}" AIBD-0814,2024-05-03,source publication,day,tested,"FDA found that the ColoSense stool RNA test detected 92.6% of colorectal cancers in 7,763 average-risk adults, more than its FIT component alone","In FDA's review of ColoSense, the stool RNA test detected 92.6% of colorectal cancers in 7,763 adults, with 85.6% specificity. It detected 41.3% of advanced precancerous lesions.",*ColoSense* stool RNA test found 25 of 27 colorectal cancers,detected 92.6% of colorectal cancers,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,ColoSense,,"Geneoscopy, Inc.",United States,Geneoscopy (CRC-PREVENT study with participants from 48 US states),United States,United States (48 contiguous states),United States,Healthcare,,Health,Prediction from data,Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,"The FIT component of ColoSense alone, in the same adults",Cancer sensitivity 77.8%,Cancer sensitivity 92.6%,measured,scaled,2026-06-09,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf23/P230001B.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. Summary of Safety and Effectiveness Data for ColoSense (P230001). Approved 2024-05-03."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf23/P230001B.pdf""}]","{""kind"": ""Official documents""}" @@ -245,16 +245,16 @@ AIBD-1088,2023-11-07,source publication,day,tested,China's CMDE finds that Aneur AIBD-1089,2023-11-03,source publication,day,tested,China's CMDE finds that AI-SONIC-TH flagged patients with thyroid nodules of TI-RADS 4 or higher with 97.07% sensitivity in the maker's trial,"In the maker's trial that China's CMDE reviewed, AI-SONIC-TH flagged patients with thyroid nodules of TI-RADS 4 or higher with 97.07% sensitivity and 90.54% specificity.",*AI-SONIC-TH* flagged suspicious thyroid nodules in China's review,97.07% sensitivity and 90.54% specificity,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,AI-SONIC-TH (甲状腺结节超声影像辅助诊断软件),Release 1 (models AI-SONIC-TH-B to AI-SONIC-TH-I),"Zhejiang Deshang Yunxing Medical Technology Co., Ltd. (浙江德尚韵兴医疗科技有限公司)",China,"Zhejiang Deshang Yunxing Medical Technology (prospective registration trial with 1,043 patients at three sites)",China,"Three trial sites that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,Targets for the lower 95% confidence bounds of per-patient sensitivity and specificity (the report does not give the values),,Sensitivity 97.07% (95%CI:95.06%,98.28%) and specificity 90.54% (95%CI:87.91%,92.64%) per patient,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1710381379631085314.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:甲状腺结节超声影像辅助诊断软件 (technical review report for the registration of thyroid nodule ultrasound image diagnosis-aid software), acceptance no. CQZ2200673, applicant 浙江德尚韵兴医疗科技有限公司. 3 November 2023. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1710381379631085314.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:甲状腺结节超声影像辅助诊断软件 (technical review report for the registration of thyroid nodule ultrasound image diagnosis-aid software), acceptance no. CQZ2200673, applicant 浙江德尚韵兴医疗科技有限公司. 3 November 2023. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1710381379631085314.pdf""}]","{""kind"": ""Official documents""}" AIBD-1066,2023-11-02,source publication,day,tested,China's CMDE finds that endoscopists using EC01-01 detected more colon polyps in the maker's trial,"In the maker's trial that China's CMDE reviewed, endoscopists using EC01-01 detected 94.7% of polyps, against 84.5% without it.",*EC01-01* helped endoscopists find more colon polyps in a trial,"detected 94.7% of polyps, against 84.5% without it",,found,Found by a regulator,controlled comparison,very low,aggregate result,false,EC01-01 (肠息肉电子下消化道图像辅助检测软件),Release V1,"Xiamen Yingnuo Medical Technology Co., Ltd. (厦门影诺医疗科技有限公司)",China,"Xiamen Yingnuo Medical Technology (prospective parallel-controlled trial with 1,020 subjects at four sites)",China,Four trial sites in China that the report does not name,China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,3,1,,,1,,organizational,2,0,"Endoscopists performing conventional colonoscopy in the parallel control group, with two experts' video review as the reference",84.5% polyp-level sensitivity,94.7% polyp-level sensitivity,measured,scaled,2025-05-22,shown,https://www.cmde.org.cn/directory/web/cmde/images/1700788508279067087.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:肠息肉电子下消化道图像辅助检测软件 (technical review report for the registration of colon polyp detection-aid software for electronic lower gastrointestinal images), acceptance no. CQZ2200605, applicant 厦门影诺医疗科技有限公司. 2 November 2023. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1700788508279067087.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:肠息肉电子下消化道图像辅助检测软件 (technical review report for the registration of colon polyp detection-aid software for electronic lower gastrointestinal images), acceptance no. CQZ2200605, applicant 厦门影诺医疗科技有限公司. 2 November 2023. Benefit-risk determination and overall opinion, CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1700788508279067087.pdf""}]","{""kind"": ""Official documents""}" AIBD-0698,2023-11-01,source publication,month,reported,UNDP's evaluation office finds that the AI system of Brazil's National Council of Justice let courts speed up high-risk cases from vulnerable groups,UNDP's evaluation office found that the AI system of Brazil's National Council of Justice let courts prioritize urgent cases. It let them prioritize high-risk cases from vulnerable groups.,Brazil's court AI helps speed up cases from vulnerable groups,,,found,Found by an evaluation office,no comparison,very low,program,false,Artificial Intelligence System of Brazil's National Council of Justice (CNJ),,"Not named in the evaluation (the system belongs to the National Council of Justice, CNJ)",,"Brazilian courts under the National Council of Justice (CNJ), with UNDP",Brazil,Brazil,Brazil,Justice administration: managing court cases,,Public sector,Prediction from data,Women and mothers,16,Upper middle income,operational_improvement,"Faster data management, decision-making",,,,4,3,,,organizational; personal,2,0,,,,,persisted,2025-03-11,not shown,https://www.unevaluation.org/get-download-file/4444,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""UNDP Independent Evaluation Office. Evaluation of the UNDP Support to Digitalization of Public Services. November 2023. The report is an independent publication of the office. Section on justice and the rule of law."", ""url"": ""https://www.unevaluation.org/get-download-file/4444""}, {""label"": ""Regulatory document"", ""value"": ""Conselho Nacional de Justiça (copy held by the Superior Tribunal de Justiça library). “Resolução Nº 615, de 11 de março de 2025.” 2025-03-11."", ""url"": ""https://www.stj.jus.br/internet_docs/biblioteca/clippinglegislacao/Res_615_2025_CNJ.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0952,2023-10-20,source publication,day,intended,Japan's Kanto District Transport Bureau approves TIER IV's automated shuttle at the GLP ALFALINK Sagamihara logistics campus as a Level 4 vehicle,On 20 October 2023 a Japanese transport bureau confirmed that TIER IV's automated operation device met the safety standards. It approved a shuttle at GLP ALFALINK Sagamihara to run with no driver.,Japan approves *AI Pilot* for a driverless shuttle at a logistics campus,,,found,Found by a regulator,,not rated,benefit created,false,Automated operation device (TIER IV's AI Pilot) on a Tajima GSM8 shuttle,,"TIER IV, Inc.",Japan,"TIER IV, Inc. (operator at GLP ALFALINK Sagamihara)",Japan,"GLP ALFALINK Sagamihara, Sagamihara, Kanagawa Prefecture, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,operational_improvement; physical_health,Enhanced productivity,,,,,,,,organizational,3,0,,,,,unknown,,not shown,https://web.archive.org/web/20260416085611id_/https://technode.global/prnasia/tier-iv-certified-in-level-4-autonomous-driving-sharing-its-design-and-process-with-partners/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Kanto District Transport Bureau. 関東で初めて自動運転車(レベル4)の認可を行いました (First approvals of Level 4 automated vehicles in Kanto). Press release, 20 October 2023, paragraphs 1 to 3 and the table's GLP ALFALINK Sagamihara column."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000304988.pdf""}]","{""kind"": ""Official documents""}" -AIBD-1213,2023-10-20,source publication,day,intended,Japan's Kanto District Transport Bureau approves a NAVYA ARMA shuttle at Haneda Innovation City as a Level 4 automated vehicle that needs no driver,On 20 October 2023 the Kanto District Transport Bureau confirmed that the ARMA's automated operation device met the safety standards. It approved the shuttle as a Level 4 vehicle on a 1.3 km loop.,NAVYA's *ARMA* may run with no driver at Haneda Innovation City,,,found,Found by a regulator,,not rated,benefit created,false,Automated operation device (自動運行装置) on a NAVYA ARMA shuttle,,NAVYA (maker of the ARMA shuttle),France,"BOLDLY Inc. (BOLDLY株式会社), the applicant and operator, inside Haneda Innovation City",Japan,"Haneda Innovation City, Ota, Tokyo, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,physical_health; operational_improvement,,,,,,,,,societal,1,0,,,,,persisted,2024-06-27,shown,https://www.anyauto.com.au/?p=16136,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 関東で初めて自動運転車(レベル4)の認可を行いました (The first approvals of Level 4 automated vehicles in Kanto). Press release, 20 October 2023."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000304988.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 関東で初めて自動運転車(レベル4)の認可を行いました (The first approvals of Level 4 automated vehicles in Kanto). Press release, 20 October 2023."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000304988.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0952,2023-10-20,source publication,day,intended,Japan's Kanto District Transport Bureau approves TIER IV's automated shuttle at the GLP ALFALINK Sagamihara logistics campus as a Level 4 vehicle,On 20 October 2023 a Japanese transport bureau confirmed that TIER IV's automated operation device met the safety standards. It approved a shuttle at GLP ALFALINK Sagamihara to run with no driver.,Japan approves *AI Pilot* for a driverless shuttle at a logistics campus,,,found,Found by a regulator,,very low,benefit created,false,Automated operation device (TIER IV's AI Pilot) on a Tajima GSM8 shuttle,,"TIER IV, Inc.",Japan,"TIER IV, Inc. (operator at GLP ALFALINK Sagamihara)",Japan,"GLP ALFALINK Sagamihara, Sagamihara, Kanagawa Prefecture, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,operational_improvement; physical_health,Enhanced productivity,,,,,,,,organizational,3,0,,,,,unknown,,not shown,https://web.archive.org/web/20260416085611id_/https://technode.global/prnasia/tier-iv-certified-in-level-4-autonomous-driving-sharing-its-design-and-process-with-partners/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Kanto District Transport Bureau. 関東で初めて自動運転車(レベル4)の認可を行いました (First approvals of Level 4 automated vehicles in Kanto). Press release, 20 October 2023, paragraphs 1 to 3 and the table's GLP ALFALINK Sagamihara column."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000304988.pdf""}]","{""kind"": ""Official documents""}" +AIBD-1213,2023-10-20,source publication,day,intended,Japan's Kanto District Transport Bureau approves a NAVYA ARMA shuttle at Haneda Innovation City as a Level 4 automated vehicle that needs no driver,On 20 October 2023 the Kanto District Transport Bureau confirmed that the ARMA's automated operation device met the safety standards. It approved the shuttle as a Level 4 vehicle on a 1.3 km loop.,NAVYA's *ARMA* may run with no driver at Haneda Innovation City,,,found,Found by a regulator,,very low,benefit created,false,Automated operation device (自動運行装置) on a NAVYA ARMA shuttle,,NAVYA (maker of the ARMA shuttle),France,"BOLDLY Inc. (BOLDLY株式会社), the applicant and operator, inside Haneda Innovation City",Japan,"Haneda Innovation City, Ota, Tokyo, Japan",Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,,11,High income,physical_health; operational_improvement,,,,,,,,,societal,1,0,,,,,persisted,2024-06-27,shown,https://www.anyauto.com.au/?p=16136,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 関東で初めて自動運転車(レベル4)の認可を行いました (The first approvals of Level 4 automated vehicles in Kanto). Press release, 20 October 2023."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000304988.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Kanto District Transport Bureau, Ministry of Land, Infrastructure, Transport and Tourism (国土交通省関東運輸局). 関東で初めて自動運転車(レベル4)の認可を行いました (The first approvals of Level 4 automated vehicles in Kanto). Press release, 20 October 2023."", ""url"": ""https://wwwtb.mlit.go.jp/kanto/content/000304988.pdf""}]","{""kind"": ""Official documents""}" AIBD-0803,2023-09-19,source publication,day,tested,FDA found that SigTuple's AI100 with Shonit matched manual microscopy on 882 blood smears in a comparison of methods at four sites,"In FDA's review of AI100 with Shonit, results on 882 blood smears agreed with manual microscopy, with 92.7% sensitivity and 95.4% specificity for abnormal white cells.",*AI100 with Shonit* matched manual microscopy on 882 blood smears,92.7% sensitivity and 95.4% specificity for abnormal white cells,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,AI100 with Shonit,,SigTuple Technologies Pvt. Ltd.,India,SigTuple (method comparison against manual microscopy at four sites),India,,,Healthcare,,Health,Images and video,Health workers,3,,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,Manual microscopy of the same 882 smears,Manual differential (reference),Sensitivity 92.7% and specificity 95.4% for abnormal white-cell results,measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/K221309.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. 510(k) substantial equivalence determination decision summary for AI100 with Shonit (K221309). Cleared 2023-09-19."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/K221309.pdf""}]","{""kind"": ""Official documents""}" AIBD-1090,2023-09-14,source publication,day,tested,China's CMDE finds that DrBrain-AD flagged hippocampal abnormalities on brain MRI with a sensitivity of 0.9328 in the maker's trial,"In the maker's trial that China's CMDE reviewed, DrBrain-AD flagged hippocampal abnormalities on brain MRI with a sensitivity of 0.9328 and a specificity of 0.9042.",*DrBrain-AD* flagged hippocampal changes on brain MRI in China's review,a sensitivity of 0.9328 and a specificity of 0.9042,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,DrBrain-AD (磁共振图像辅助评估软件),Release 1.0 (full version 1.0.0.01),"Shenzhen Yiwei Medical Technology Co., Ltd. (深圳市铱硙医疗科技有限公司)",China,Shenzhen Yiwei Medical Technology (registration trial with 514 patients at three hospitals),China,"Xuanwu Hospital of Capital Medical University, the Second Xiangya Hospital of Central South University and the First Affiliated Hospital of Wenzhou Medical University, China",China,Healthcare,,Health,Images and video,Health workers; Patients,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,Targets of 85% for sensitivity and 80% for specificity for the lower 95% confidence bounds,Targets of 85% and 80%,Sensitivity 0.9328 (0.9020,0.9637) and specificity 0.9042 (0.8685,0.9399),measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1710381691379018965.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:磁共振图像辅助评估软件 (technical review report for the registration of magnetic resonance image assessment-aid software), acceptance no. CQZ2201033, applicant 深圳市铱硙医疗科技有限公司. 14 September 2023. Section 三、临床评价概述 (clinical evaluation)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1710381691379018965.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:磁共振图像辅助评估软件 (technical review report for the registration of magnetic resonance image assessment-aid software), acceptance no. CQZ2201033, applicant 深圳市铱硙医疗科技有限公司. 14 September 2023. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment of benefit and risk."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1710381691379018965.pdf""}]","{""kind"": ""Official documents""}" AIBD-0631,2023-08-03,source publication,day,tested,FDA De Novo review of Viz HCM: AI detects hypertrophic cardiomyopathy on 12-lead ECGs from three Massachusetts hospitals,"In FDA's review of Viz HCM, the software detected hypertrophic cardiomyopathy on 12-lead ECGs from three hospitals with 68.4% sensitivity. Specificity was 99.1%.",*Viz HCM* picked out a hard-to-diagnose heart disease from ECGs,a sensitivity of 68.4% and a specificity of 99.1%,,found,Found by a regulator,no comparison,very low,aggregate result,false,Viz HCM,,"Viz.ai, Inc.",United States,Viz.ai (retrospective study on ECGs from three Massachusetts hospitals),United States,"Boston area and Salem, Massachusetts, United States",United States,Healthcare,,Health,Prediction from data,Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,,,,,persisted,2026-03-23,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230003.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for Viz HCM (DEN230003): decision summary. Granted 2023-08-03."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230003.pdf""}]","{""kind"": ""Official documents""}" AIBD-1067,2023-07-11,source publication,day,tested,"China's CMDE finds that with EASYASPECT, doctors' scores of stroke CT scans agreed better with an expert panel in the maker's trial","In the maker's reader trial that China's CMDE reviewed, doctors' judgments of stroke CT agreed with an expert panel at a kappa of 0.80 with EASYASPECT, against 0.55 without it.",*EASYASPECT* brought doctors' scores of stroke CT closer to experts',"a kappa of 0.80 with EASYASPECT, against 0.55 without it",,found,Found by a regulator,controlled comparison,very low,aggregate result,false,EASYASPECT (缺血性卒中CT图像辅助评估软件),Release 1.0 (full version 1.0.0.007),"Yukun (Beijing) Network Technology Co., Ltd. (语坤(北京)网络科技有限公司)",China,Yukun (Beijing) Network Technology (retrospective reader trial at the Second Affiliated Hospital of Nanchang University and Hwa Mei Hospital of the University of Chinese Academy of Sciences in Ningbo),China,"Ningbo, Zhejiang and Nanchang, Jiangxi, China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,2,1,,,1,,organizational,1,0,"The same eight doctors scoring the same cases without the software, with an expert panel as the reference",Kappa 0.55 between doctors without the software and the expert panel,Kappa 0.80 between doctors with the software and the expert panel (lower 95% confidence bound 0.79),measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1743664671250013909.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:缺血性卒中CT图像辅助评估软件 (technical review report for the registration of ischemic stroke CT image assessment-aid software), acceptance no. CQZ2201543, applicant 语坤(北京)网络科技有限公司. 11 July 2023. Section 三、临床评价概述 (clinical evaluation), pp. 7-8."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1743664671250013909.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:缺血性卒中CT图像辅助评估软件 (technical review report for the registration of ischemic stroke CT image assessment-aid software), acceptance no. CQZ2201543, applicant 语坤(北京)网络科技有限公司. 11 July 2023. Section 综合评价意见 (concluding opinion), CMDE's own judgment."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1743664671250013909.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0736,2023-07-05,source publication,day,tested,NICE finds that AI lung nodule software with clinician review could detect more cancers at slightly lower cost in targeted lung cancer screening,"In a model by NICE's assessors, AI nodule software with clinician review in targeted lung screening found more people needing follow-up at slightly lower cost. It found slightly more cancers.",Lung screening centers may use AI nodule detection under NICE guidance,"In all nine studies of reading time, review was faster with the software",,found,Found by an assessment body,,very low,aggregate result,false,"AI-derived computer-aided detection software for lung nodules: AI-Rad Companion Chest CT, AVIEW LCS+, ClearRead CT, contextflow SEARCH Lung CT, InferRead CT Lung, Lung AI, qCT-Lung, SenseCare-Lung Pro, Veolity, Veye Lung Nodules and VUNO Med-LungCT AI",,"Siemens Healthineers, Coreline Soft, Riverain Technologies, contextflow, Infervision, Arterys, Qure.ai, SenseTime, MeVis, Aidence and VUNO","Multiple: Germany, South Korea, United States, Austria, China, India, Netherlands",Screening centers in the studies reviewed by NICE and NHS targeted lung cancer screening centers in England during evidence generation,Multiple,"England, United Kingdom (guidance), with studies in several countries",United Kingdom,Healthcare (lung cancer screening),,Health,Images and video,Health workers,3,High income,physical_health; operational_improvement,Medical diagnosis,,,1,,,,,organizational,3,0,Clinician review of screening CT scans without the software,,More sensitive but less specific in nearly all studies,measured,persisted,,shown,https://www.iqwig.de/download/t24-04_ai-unterstuetzte-diagnose-bei-lungenkrebs_themencheck-bericht_v1-0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI)-derived computer-aided detection (CAD) software for detecting and measuring lung nodules in CT scan images. HealthTech guidance HTG687 (formerly DG55), section 3: Committee discussion. 5 July 2023."", ""url"": ""https://www.nice.org.uk/guidance/htg687/chapter/3-Committee-discussion""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HTG687, section 1: Recommendations. 5 July 2023."", ""url"": ""https://www.nice.org.uk/guidance/htg687/chapter/1-Recommendations""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG), ThemenCheck Medizin (2026). Lungenkrebs: Kann Künstliche Intelligenz die Auswertung von Röntgen- oder CT-Bildern der Lunge unterstützen? ThemenCheck-Bericht T24-04, Version 1.0, Kurzfassung, Fazit (pp. 6-7); included studies in Tabelle 11 (Lee 2025: LuCAS-Plus; Lo 2018: ClearRead CT)."", ""url"": ""https://www.iqwig.de/download/t24-04_ai-unterstuetzte-diagnose-bei-lungenkrebs_themencheck-bericht_v1-0.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0736,2023-07-05,source publication,day,tested,NICE finds that AI lung nodule software with clinician review could detect more cancers at slightly lower cost in targeted lung cancer screening,"In a model by NICE's assessors, AI nodule software with clinician review in targeted lung screening found more people needing follow-up at slightly lower cost. It found slightly more cancers.",Lung screening centers may use AI nodule detection under NICE guidance,"In all nine studies of reading time, review was faster with the software",,found,Found by an assessment body,controlled comparison,very low,aggregate result,false,"AI-derived computer-aided detection software for lung nodules: AI-Rad Companion Chest CT, AVIEW LCS+, ClearRead CT, contextflow SEARCH Lung CT, InferRead CT Lung, Lung AI, qCT-Lung, SenseCare-Lung Pro, Veolity, Veye Lung Nodules and VUNO Med-LungCT AI",,"Siemens Healthineers, Coreline Soft, Riverain Technologies, contextflow, Infervision, Arterys, Qure.ai, SenseTime, MeVis, Aidence and VUNO","Multiple: Germany, South Korea, United States, Austria, China, India, Netherlands",Screening centers in the studies reviewed by NICE and NHS targeted lung cancer screening centers in England during evidence generation,Multiple,"England, United Kingdom (guidance), with studies in several countries",United Kingdom,Healthcare (lung cancer screening),,Health,Images and video,Health workers,3,High income,physical_health; operational_improvement,Medical diagnosis,,,1,,,,,organizational,3,0,Clinician review of screening CT scans without the software,,More sensitive but less specific in nearly all studies,measured,persisted,,shown,https://www.iqwig.de/download/t24-04_ai-unterstuetzte-diagnose-bei-lungenkrebs_themencheck-bericht_v1-0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI)-derived computer-aided detection (CAD) software for detecting and measuring lung nodules in CT scan images. HealthTech guidance HTG687 (formerly DG55), section 3: Committee discussion. 5 July 2023."", ""url"": ""https://www.nice.org.uk/guidance/htg687/chapter/3-Committee-discussion""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HTG687, section 1: Recommendations. 5 July 2023."", ""url"": ""https://www.nice.org.uk/guidance/htg687/chapter/1-Recommendations""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG), ThemenCheck Medizin (2026). Lungenkrebs: Kann Künstliche Intelligenz die Auswertung von Röntgen- oder CT-Bildern der Lunge unterstützen? ThemenCheck-Bericht T24-04, Version 1.0, Kurzfassung, Fazit (pp. 6-7); included studies in Tabelle 11 (Lee 2025: LuCAS-Plus; Lo 2018: ClearRead CT)."", ""url"": ""https://www.iqwig.de/download/t24-04_ai-unterstuetzte-diagnose-bei-lungenkrebs_themencheck-bericht_v1-0.pdf""}]","{""kind"": ""Official documents""}" AIBD-1140,2023-07-01,source publication,month,tested,"The EU's ECIBC guideline group suggests AI-supported double reading of mammograms, finding that it may detect one more cancer per 1,000 women","In July 2023, the ECIBC guideline group suggested AI-supported double reading of screening mammograms. It found that AI support may detect one more breast cancer for every 1,000 women screened.",EU guideline group backs *AI-supported double reading* of mammograms,"one more breast cancer for every 1,000 women screened",,found,Found by an assessment body,tested against a standard,very low,aggregate result,false,"AI support for double reading of screening mammograms: three commercial algorithms kept anonymous in the study behind the estimate, with Transpara, Lunit INSIGHT MMG, Mia and a DeepHealth model in the other studies reviewed",,"Three vendors that asked to remain anonymous in the study behind the estimate, with ScreenPoint Medical (Transpara), Lunit (INSIGHT MMG), Kheiron Medical Technologies (Mia) and DeepHealth in the other studies reviewed","Multiple: Netherlands, South Korea, United Kingdom, United States","Screening services in the studies that the ECIBC guideline group reviewed, with Karolinska University Hospital's archived examinations behind the estimate and organized screening programs in the EU as intended users",Multiple,"European Union (guideline), with the study behind the estimate in Stockholm, Sweden",Sweden,Healthcare,,Health,Images and video,Women and mothers,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,3,0,"Double reading without AI support, pooled from six studies that the guideline group reviewed",,one more breast cancer detected for every 1 000 women screened,measured,persisted,2026-11-06,shown,https://cancer-screening-and-care.jrc.ec.europa.eu/en/ecibc/european-breast-cancer-guidelines?topic=64&usertype=60,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""European Commission Initiative on Breast Cancer (ECIBC), Guidelines Development Group, European Commission Joint Research Centre. European guidelines on breast cancer screening and diagnosis, printed overview of the recommendations. Breast cancer screening, 4. Use of artificial intelligence, recommendation 4.2. Issued July 2023."", ""url"": ""https://cancer-screening-and-care.jrc.ec.europa.eu/en/ecibc/recommendations/print""}, {""label"": ""Assessment report"", ""value"": ""European Commission Initiative on Breast Cancer (ECIBC), Guidelines Development Group. European guidelines on breast cancer screening and diagnosis, 4. Use of artificial intelligence, recommendations 4.1 and 4.2 with their justifications. Issued July 2023. Supporting material: GRADE Evidence-to-Decision Framework for recommendation 4.2 (https://dbep.ec.gdt.evidenceprime.com/profile/PNED9kXB-h0), which cites Salim et al. 2020 for double reading with AI support and pools that study with Dahlblom et al. 2022, Dembrower et al. 2023, Larsen et al. 2022, Marinovich et al. 2023 and Sharma et al. 2023 for double reading without it."", ""url"": ""https://cancer-screening-and-care.jrc.ec.europa.eu/en/ecibc/european-breast-cancer-guidelines?topic=64&usertype=60""}]","{""kind"": ""Official documents""}" AIBD-0776,2023-06-29,source publication,day,tested,FDA's De Novo review of KidneyIntelX.dkd shows that its risk levels separated patients with diabetic kidney disease by their five-year risk of serious decline,"In FDA's review, adults whom KidneyIntelX.dkd placed at low risk had an estimated 5.9% rate of serious kidney decline over five years. The rate was 21.5% at moderate risk. At high risk it was 66.9%.",*KidneyIntelX.dkd* sorted diabetic patients by risk to their kidneys,"5.9%, 21.5% and 66.9%",,found,Found by a regulator,no comparison,very low,aggregate result,false,KidneyIntelX.dkd,,"Renalytix AI, Inc.",United States,Renalytix (retrospective validation on biobanked samples from 657 adults),United States,A single biorepository drawing on 10 clinical sites (not named in the decision summary),,Healthcare,,Health,Prediction from data,Patients; Health workers,3,,physical_health,Medical diagnosis,,,,3,,,,organizational,5,0,,,,,persisted,2026-04-22,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200052.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. Evaluation of automatic class III designation for KidneyIntelX.dkd (DEN200052): decision summary. Granted 2023-06-29."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200052.pdf""}, {""label"": ""Assessment report"", ""value"": ""Wellpoint Federal (formerly National Government Services), Medicare Administrative Contractor, Jurisdictions 6 and K (2024). LCD L39726: KidneyIntelX and KidneyIntelX.dkd Testing (original effective date 1 August 2024, revision effective 1 April 2026), Coverage Indications; Summary of Evidence; Analysis of Evidence (Rationale for Determination)."", ""url"": ""https://www.cms.gov/medicare-coverage-database/view/lcd.aspx?lcdid=39726""}, {""label"": ""Assessment report"", ""value"": ""Anthem Blue Cross and Blue Shield (Elevance Health), Medical Policy LAB.00041 (2026). Medical Policy LAB.00041: Machine Learning Derived Probability Score for Rapid Kidney Function Decline (last review 14 May 2026, published 1 July 2026), Position Statement; Summary for Members and Families ('What the Studies Show'); Rationale."", ""url"": ""https://www.anthem.com/dam/medpolicies/abcbs/active/policies/mp_pw_e001134.html""}]","{""kind"": ""Official documents""}" -AIBD-1068,2023-05-18,source publication,day,tested,China's CMDE finds that Tencent's MY-CJ-01 software raised the detection of polyps on a first pass in the maker's randomized colonoscopy trial,"China's CMDE found that first-pass polyp detection was superior in group B of Tencent's randomized MY-CJ-01 trial. The polyp miss rate was 23.39%, against 44.67% in group A.",*MY-CJ-01* raised detection of polyps on a first pass in Tencent's trial,"The polyp miss rate was 23.39%, against 44.67% in group A",,found,Found by a regulator,,very low,aggregate result,false,MY-CJ-01 (结肠息肉电子内窥镜图像辅助检测软件),Release V2 (full version V2.0.0_build001),"Tencent Healthcare (Shenzhen) Co., Ltd. (腾讯医疗健康(深圳)有限公司)",China,Tencent Healthcare (Shenzhen) (prospective randomized trial with 296 patients at two centers),China,"Two trial centers that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,3,1,,,1,,organizational,2,0,Group A of the same randomized trial (the report does not say how each group used the software),Polyp miss rate 44.67% (95%CI:39.07%~50.42%) in group A,Polyp miss rate 23.39% (95%CI:18.92%~28.54%) in group B,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1686295172757009730.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:结肠息肉电子内窥镜图像辅助检测软件 (technical review report for the registration of colon polyp electronic endoscope image detection-aid software), acceptance no. CQZ2200455, applicant 腾讯医疗健康(深圳)有限公司. 18 May 2023. Section 三、临床评价概述 (clinical evaluation), pp. 6-7."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1686295172757009730.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:结肠息肉电子内窥镜图像辅助检测软件 (technical review report for the registration of colon polyp electronic endoscope image detection-aid software), acceptance no. CQZ2200455, applicant 腾讯医疗健康(深圳)有限公司. 18 May 2023. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1686295172757009730.pdf""}, {""label"": ""Regulatory document"", ""value"": ""National Medical Products Administration (NMPA), China. Computer-aided Detection Software for Electronic Endoscope for Colon Polyps Approved for Marketing. NMPA English website, 1 June 2023."", ""url"": ""https://english.nmpa.gov.cn/2023-06/01/c_940345.htm""}]","{""kind"": ""Official documents""}" +AIBD-1068,2023-05-18,source publication,day,tested,China's CMDE finds that Tencent's MY-CJ-01 software raised the detection of polyps on a first pass in the maker's randomized colonoscopy trial,"China's CMDE found that first-pass polyp detection was superior in group B of Tencent's randomized MY-CJ-01 trial. The polyp miss rate was 23.39%, against 44.67% in group A.",*MY-CJ-01* raised detection of polyps on a first pass in Tencent's trial,"The polyp miss rate was 23.39%, against 44.67% in group A",,found,Found by a regulator,controlled comparison,very low,aggregate result,false,MY-CJ-01 (结肠息肉电子内窥镜图像辅助检测软件),Release V2 (full version V2.0.0_build001),"Tencent Healthcare (Shenzhen) Co., Ltd. (腾讯医疗健康(深圳)有限公司)",China,Tencent Healthcare (Shenzhen) (prospective randomized trial with 296 patients at two centers),China,"Two trial centers that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,physical_health,Medical diagnosis,,3,1,,,1,,organizational,2,0,Group A of the same randomized trial (the report does not say how each group used the software),Polyp miss rate 44.67% (95%CI:39.07%~50.42%) in group A,Polyp miss rate 23.39% (95%CI:18.92%~28.54%) in group B,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1686295172757009730.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:结肠息肉电子内窥镜图像辅助检测软件 (technical review report for the registration of colon polyp electronic endoscope image detection-aid software), acceptance no. CQZ2200455, applicant 腾讯医疗健康(深圳)有限公司. 18 May 2023. Section 三、临床评价概述 (clinical evaluation), pp. 6-7."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1686295172757009730.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:结肠息肉电子内窥镜图像辅助检测软件 (technical review report for the registration of colon polyp electronic endoscope image detection-aid software), acceptance no. CQZ2200455, applicant 腾讯医疗健康(深圳)有限公司. 18 May 2023. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1686295172757009730.pdf""}, {""label"": ""Regulatory document"", ""value"": ""National Medical Products Administration (NMPA), China. Computer-aided Detection Software for Electronic Endoscope for Colon Polyps Approved for Marketing. NMPA English website, 1 June 2023."", ""url"": ""https://english.nmpa.gov.cn/2023-06/01/c_940345.htm""}]","{""kind"": ""Official documents""}" AIBD-2306,2023-03-31,source publication,day,tested,FDA De Novo review of the Masimo SafetyNet Opioid System: pattern recognition matched a fixed-threshold oximeter's sensitivity for opioid-induced respiratory depression with fewer non-actionable alarms,"In data from 641 opioid users that FDA reviewed, Masimo's Opioid Halo software matched a fixed-threshold oximeter's sensitivity for respiratory depression and cut needless alarms by 76% to 96.7%.",*Opioid Halo* cut needless oxygen alarms for people who take opioids,96.7%,,found,Found by a regulator,tested against a standard,very low,program,false,Masimo SafetyNet Opioid System (Opioid Halo),,Masimo Corporation,United States,"Masimo (analysis of recordings from people who took opioids at home, in hospital or illicitly)",United States,United States,United States,Healthcare (opioid safety),,Health,Prediction from data,Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,3,0,A pulse oximeter with a fixed alarm threshold on the same recordings,Sensitivity 99.2% and specificity 80.6% overall,"Sensitivity 100.0% and specificity 93.9% at the first alarm level, with 76% fewer non-actionable alarms",measured,persisted,2025-02-06,shown,https://www.masimo.de/siteassets/us/documents/pdf/plm-13880b-brochure-opioid-halo---hcp-us.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for the Masimo SafetyNet Opioid System (DEN200011): decision summary. Granted 2023-03-31. The summary is a scanned document, and the quotations follow its text as a machine reads it."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200011.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for the Masimo SafetyNet Opioid System (DEN200011): decision summary. Granted 2023-03-31. The summary is a scanned document, and the quotations follow its text as a machine reads it."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200011.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for the Masimo SafetyNet Opioid System (DEN200011): decision summary. Granted 2023-03-31. The summary is a scanned document, and the quotations follow its text as a machine reads it."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200011.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for the Masimo SafetyNet Opioid System (DEN200011): decision summary. Granted 2023-03-31. The summary is a scanned document, and the quotations follow its text as a machine reads it."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200011.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for the Masimo SafetyNet Opioid System (DEN200011): decision summary. Granted 2023-03-31. The summary is a scanned document, and the quotations follow its text as a machine reads it."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200011.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for the Masimo SafetyNet Opioid System (DEN200011): decision summary. Granted 2023-03-31. The summary is a scanned document, and the quotations follow its text as a machine reads it."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200011.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo classification request for the Masimo SafetyNet Opioid System (DEN200011): decision summary. Granted 2023-03-31. The summary is a scanned document, and the quotations follow its text as a machine reads it."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200011.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration. Artificial Intelligence-Enabled Medical Devices. List current as of 4 September 2026. The quotation is the device's name in the list. The list contains no benefit finding."", ""url"": ""https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-enabled-medical-devices""}]","{""kind"": ""Official documents""}" AIBD-1070,2023-03-03,source publication,day,tested,"China's CMDE finds that pathologists read cervical cytology slides faster with CCAA.91360, with accuracy no worse, in the maker's trial","In the maker's trial that China's CMDE reviewed, pathologists took a mean 64.12 seconds per cervical cytology slide with CCAA.91360, against 333.49 seconds reading digital slides alone.",*CCAA.91360* cut the time to read cervical cytology slides,"a mean 64.12 seconds per cervical cytology slide with CCAA.91360, against 333.49 seconds",,found,Found by a regulator,controlled comparison,very low,aggregate result,false,CCAA.91360 (宫颈细胞学数字病理图像计算机辅助分析软件),Release V2,"91360 Medical Technology (Nanjing) Co., Ltd. (玖壹叁陆零医学科技南京有限公司)",China,91360 Medical Technology (Nanjing) (registration trial at the First Affiliated Hospital of Nanjing Medical University and the Affiliated Drum Tower Hospital of Nanjing University Medical School),China,"Nanjing, Jiangsu, China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,operational_improvement,Enhanced productivity,,,1,,,1,,organizational,1,0,Digital reading of the same slides without the software,Mean diagnosis time 333.49 seconds,Mean diagnosis time 64.12 seconds,measured,unknown,,shown,https://www.cmde.org.cn/directory/web/cmde/images/1710380539480034481.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:宫颈细胞学数字病理图像计算机辅助分析软件 (technical review report for the registration of cervical cytology digital pathology image computer-aided analysis software), acceptance no. CQZ2101339, applicant 玖壹叁陆零医学科技南京有限公司. 3 March 2023. Section 三、临床评价概述 (clinical evaluation), pp. 7-8."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1710380539480034481.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:宫颈细胞学数字病理图像计算机辅助分析软件 (technical review report for the registration of cervical cytology digital pathology image computer-aided analysis software), acceptance no. CQZ2101339, applicant 玖壹叁陆零医学科技南京有限公司. 3 March 2023. Section 四、产品受益风险判定 (benefit-risk determination), CMDE's own judgment."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/1710380539480034481.pdf""}]","{""kind"": ""Official documents""}" AIBD-0778,2023-02-24,source publication,day,tested,FDA finds that Caption Health's AutoEF met its goal for accuracy when it estimated ejection fraction on echocardiograms from 186 patients,"In FDA's review of Caption Interpretation AutoEF, its estimates of ejection fraction on 186 patient studies differed from expert readings by an RMSD of 7.21. The goal was below 9.2%.",*AutoEF* met its goal for accuracy in measuring how well the heart pumps,an RMSD of 7.21,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Caption Interpretation Automated Ejection Fraction Software (AutoEF),,"Caption Health, Inc.",United States,Caption Health (retrospective validation on 186 patient studies from three US sites),United States,"Minneapolis Heart Institute, Duke University and Northwestern University, United States",United States,Healthcare,,Health,Images and video,Health workers,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,Pre-specified acceptance criterion for the root mean square deviation from expert measurements,Goal of RMSD below 9.2%,RMSD 7.21 for the best available view,measured,scaled,2024-04-11,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN220063.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo Classification Request for Caption Interpretation Automated Ejection Fraction Software (DEN220063): decision summary. Granted 2023-02-24."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN220063.pdf""}]","{""kind"": ""Official documents""}" @@ -275,7 +275,7 @@ AIBD-0624,2021-09-21,source publication,day,tested,FDA De Novo review of Paige P AIBD-1123,2021-09-13,source publication,day,mixed,"Scotland's SHTG advises that HeartFlow FFRCT may reduce invasive coronary angiography, on supporting evidence of limited quality",SHTG advised NHSScotland in September 2021 that HeartFlow FFRCT may reduce the need for invasive coronary angiography. The test may also save costs. SHTG judged the evidence limited in quality.,Scotland backs *HeartFlow FFRCT* as an option to avoid invasive tests,,,found,Found by an assessment body,controlled comparison,low,aggregate result,false,HeartFlow FFRCT,,"HeartFlow, Inc.",United States,"NHSScotland health boards, with a pilot at NHS Western Isles from December 2020",United Kingdom,"Scotland, United Kingdom",United Kingdom,Healthcare,,Health,Images and video,Health workers; Patients,3,High income,physical_health,Medical diagnosis,,,,,2,,,personal,6,0,Standard care for assessing coronary artery disease after CT coronary angiography,,,measured,persisted,2026-05-01,not shown,https://admin.shtg.scot/media/2087/20210714-heartflow-adaptation-v20.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Scottish Health Technologies Group, Healthcare Improvement Scotland. HeartFlow FFRCT for estimating fractional flow reserve from coronary CT angiography: Recommendation for NHSScotland (SHTG Adaptation of NICE guidance). Published 13 September 2021. Archived copy of 27 April 2025."", ""url"": ""https://web.archive.org/web/20250427064346id_/https://shtg.scot/our-advice/heartflow-ffrct-for-estimating-fractional-flow-reserve-from-coronary-ct-angiography/""}, {""label"": ""Assessment report"", ""value"": ""Scottish Health Technologies Group. SHTG Adaptation: HeartFlow FFRCT for estimating fractional flow reserve from coronary CT angiography. 3 September 2021, Scottish Health Technology Council considerations."", ""url"": ""https://admin.shtg.scot/media/2087/20210714-heartflow-adaptation-v20.pdf""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HeartFlow FFRCT for estimating fractional flow reserve from coronary CT angiography. HealthTech guidance HTG429 (formerly MTG32), section 1: Recommendations. 13 February 2017, last updated 19 May 2021."", ""url"": ""https://www.nice.org.uk/guidance/htg429/chapter/1-Recommendations""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HeartFlow FFRCT for estimating fractional flow reserve from coronary CT angiography. HealthTech guidance HTG429 (formerly MTG32), section 1: Recommendations. 13 February 2017, last updated 19 May 2021."", ""url"": ""https://www.nice.org.uk/guidance/htg429/chapter/1-Recommendations""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HeartFlow FFRCT for estimating fractional flow reserve from coronary CT angiography. HealthTech guidance HTG429 (formerly MTG32), section 6: Conclusions. 13 February 2017, last updated 19 May 2021."", ""url"": ""https://www.nice.org.uk/guidance/htg429/chapter/6-Conclusions""}, {""label"": ""Assessment report"", ""value"": ""Evaluación de Tecnologías Sanitarias de Andalucía (AETSA). Sistemas de estimación del flujo coronario y de la perfusión miocárdica mediante software de análisis de angiografías. Informe de Evaluación de Tecnologías Sanitarias, Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del Sistema Nacional de Salud. Ministerio de Sanidad, 2025. The quotations are in Spanish."", ""url"": ""https://www.aetsa.org/download/publicaciones/22_2025_AETSA_-Sistema-de-estimacion-flujo-coronario_web.pdf""}, {""label"": ""Assessment report"", ""value"": ""Evaluación de Tecnologías Sanitarias de Andalucía (AETSA). Sistemas de estimación del flujo coronario y de la perfusión miocárdica mediante software de análisis de angiografías. Informe de Evaluación de Tecnologías Sanitarias, Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del Sistema Nacional de Salud. Ministerio de Sanidad, 2025. The quotations are in Spanish."", ""url"": ""https://www.aetsa.org/download/publicaciones/22_2025_AETSA_-Sistema-de-estimacion-flujo-coronario_web.pdf""}]","{""kind"": ""Official documents""}" AIBD-1042,2021-09-01,source publication,month,reported,"GAFILAT assessors find that SIMONE, among other systems, gives Chile's financial intelligence unit timely, good-quality analysis","GAFILAT assessors found that SIMONE, among other systems, gives Chile's financial intelligence unit timely, good-quality analysis. They found that it checks UN sanctions lists every 5 minutes.",*SIMONE* gives Chile's financial intelligence unit timely analysis,more than 60 million records and transactions,,found,Found by an assessment body,no comparison,very low,program,false,SIMONE (Sistema de Monitoreo Estratégico),,"Unidad de Análisis Financiero (UAF), strategic intelligence area",Chile,"Unidad de Análisis Financiero (UAF), Chile's financial intelligence unit",Chile,Chile (national),Chile,Anti-money laundering: financial intelligence,,Finance,Prediction from data,Public servants,16,High income,operational_improvement,"Faster data management, decision-making",,1,,,3,5,,organizational,1,0,,,,,unknown,,shown,https://biblioteca.gafilat.org/wp-content/uploads/2024/07/IEM-Chile.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""GAFILAT. Informe de Evaluación Mutua de la República de Chile (cuarta ronda). September 2021, adopted at the XLIII GAFILAT Plenary of July 2021. Chapter 3, main findings for Immediate Outcome 6, paras 164 and 179, and para 371."", ""url"": ""https://biblioteca.gafilat.org/wp-content/uploads/2024/07/IEM-Chile.pdf""}]","{""kind"": ""Official documents""}" AIBD-0824,2021-08-31,source publication,day,tested,The FCC certifies CaptionCall's fully automatic captions after its test lab measured fewer word errors than in captions made with human assistants,"In the FCC's National Test Lab, CaptionCall's automatic captions had median error rates of 2.9 to 14.5 across five scenarios. Human-assisted providers had 8.9 to 19.5 on average.",*CaptionCall*'s automatic captions beat assisted ones in the FCC's lab,median error rates of 2.9 to 14.5,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,CaptionCall ASR-only (fully automatic) Internet Protocol Captioned Telephone Service (IP CTS),,"CaptionCall, LLC (with Sorenson Communications, LLC), on a speech recognition platform from an unnamed vendor",United States,"CaptionCall, LLC",United States,United States,United States,Telecommunications relay services / accessibility for people with hearing loss,,Accessibility,Speech and sound,People with disabilities,10,High income,accessibility,,,,4,4,,,,personal,2,0,Median error rates of CA-assisted providers on the same five call scenarios,8.9 to 19.5 on average,2.9 to 14.5,measured,persisted,2025-06-24,shown,https://docs.fcc.gov/public/attachments/DA-25-542A1.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Federal Communications Commission, Consumer and Governmental Affairs Bureau. Order, DA 21-1080, CG Docket No. 03-123 (conditional certification of CaptionCall, LLC to provide IP CTS on a fully automatic basis). Adopted and released 31 August 2021."", ""url"": ""https://docs.fcc.gov/public/attachments/DA-21-1080A1.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Federal Communications Commission, Consumer and Governmental Affairs Bureau. Order, DA 21-1080, CG Docket No. 03-123 (conditional certification of CaptionCall, LLC to provide IP CTS on a fully automatic basis). Adopted and released 31 August 2021."", ""url"": ""https://docs.fcc.gov/public/attachments/DA-21-1080A1.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Federal Communications Commission, Consumer and Governmental Affairs Bureau (2025). Public Notice, DA 25-542, CG Docket No. 03-123: Grant of Certification for CaptionCall, LLC, to Provide Internet Protocol Captioned Telephone Service under Ownership of Ariel GP Holdco, LLC, p. 1 and the paragraph on consent decrees (p. 3-4)."", ""url"": ""https://docs.fcc.gov/public/attachments/DA-25-542A1.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0993,2021-08-10,source publication,day,intended,A California federal court held that Southwest Airlines' keyword searches with technology-assisted review were consistent with a reasonable inquiry,In August 2021 a federal court in California held that Southwest Airlines' use of keyword searches with technology-assisted review was consistent with a reasonable inquiry in a class action.,Court backs Southwest's *technology-assisted review* with keywords,"over 47,000 responsive documents",,found,Found by a court,,not rated,benefit created,false,"Technology-assisted review, used with keyword searches",,Not named in the order,,Southwest Airlines Co. (defendant),United States,"California, United States",United States,Legal services / e-discovery,,Science,Text and language,,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,,,,,unknown,,not shown,https://storage.courtlistener.com/recap/gov.uscourts.cand.336694/gov.uscourts.cand.336694.131.0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Northern District of California (Judge Phyllis J. Hamilton). Huntsman v. Southwest Airlines Co., No. 19-cv-00083-PJH, Discovery Order (Doc. 131), 10 August 2021, pp. 3-5. CourtListener RECAP copy."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.cand.336694/gov.uscourts.cand.336694.131.0.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0993,2021-08-10,source publication,day,intended,A California federal court held that Southwest Airlines' keyword searches with technology-assisted review were consistent with a reasonable inquiry,In August 2021 a federal court in California held that Southwest Airlines' use of keyword searches with technology-assisted review was consistent with a reasonable inquiry in a class action.,Court backs Southwest's *technology-assisted review* with keywords,"over 47,000 responsive documents",,found,Found by a court,,very low,benefit created,false,"Technology-assisted review, used with keyword searches",,Not named in the order,,Southwest Airlines Co. (defendant),United States,"California, United States",United States,Legal services / e-discovery,,Science,Text and language,,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,,,,,unknown,,not shown,https://storage.courtlistener.com/recap/gov.uscourts.cand.336694/gov.uscourts.cand.336694.131.0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Northern District of California (Judge Phyllis J. Hamilton). Huntsman v. Southwest Airlines Co., No. 19-cv-00083-PJH, Discovery Order (Doc. 131), 10 August 2021, pp. 3-5. CourtListener RECAP copy."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.cand.336694/gov.uscourts.cand.336694.131.0.pdf""}]","{""kind"": ""Official documents""}" AIBD-0780,2021-05-01,event,month,tested,FDA finds that ScanNav Anatomy Peripheral Nerve Block helped 19 of 30 anesthesiologists obtain the correct ultrasound view in a study,"In a study that FDA reviewed, 63% (19/30) of anesthesiologists judged that ScanNav Anatomy Peripheral Nerve Block helped them obtain the correct ultrasound view before a block.",*ScanNav Anatomy PNB* helped anesthesiologists find the right view,63% (19/30) of anesthesiologists,,found,Found by a regulator,no comparison,very low,aggregate result,false,ScanNav Anatomy Peripheral Nerve Block,,Intelligent Ultrasound Limited,United Kingdom,Intelligent Ultrasound (study with 15 expert and 15 trainee anesthesiologists at Oregon Health & Science University),United States,"Portland, Oregon, United States",United States,Healthcare,,Health,Images and video,Health workers,3,High income,operational_improvement,Value above replacement,,2,1,,,1,,organizational,2,0,,,,,persisted,2023-10-06,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN220024.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo Classification Request for ScanNav Anatomy Peripheral Nerve Block (DEN220024): decision summary. Granted 2022-10-18."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN220024.pdf""}]","{""kind"": ""Official documents""}" AIBD-1000,2021-04-13,source publication,day,tested,The FCC certifies ClearCaptions' automatic captions after its test lab measured fewer word errors than in human-assisted captions,"In the FCC's test lab, ClearCaptions' automatic captions had median word error rates of 3.5% to 14.8%, against 8.9% to 19.5% for assisted providers. The FCC fully certified the service in 2023.",*ClearCaptions*' automatic captions beat assisted ones in the FCC's lab,median word error rates of 3.5% to 14.8%,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,ClearCaptions fully automatic (ASR-only) Internet Protocol Captioned Telephone Service (IP CTS),,"ClearCaptions, LLC, on the platform of a speech-to-text vendor that the FCC does not name",,"ClearCaptions, LLC",,United States,United States,Telecommunications relay services / accessibility for people with hearing loss,,Accessibility,Speech and sound,People with disabilities,10,High income,accessibility,,,,4,4,,5,,personal,3,0,"Median word error rates of CA-assisted providers, on average, on the same call scenarios",8.9% to 19.5%,3.5% to 14.8%,measured,persisted,2023-04-04,shown,https://docs.fcc.gov/public/attachments/DA-21-419A1.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Federal Communications Commission, Consumer and Governmental Affairs Bureau. Memorandum Opinion and Order, DA 21-419, CG Docket No. 03-123 (conditional certification of ClearCaptions, LLC to provide IP CTS on a fully automatic basis). Adopted and released 13 April 2021."", ""url"": ""https://docs.fcc.gov/public/attachments/DA-21-419A1.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Federal Communications Commission, Consumer and Governmental Affairs Bureau. Order, DA 23-291, CG Docket No. 03-123 (full certification of ClearCaptions, LLC to provide IP CTS). Adopted and released 4 April 2023."", ""url"": ""https://docs.fcc.gov/public/attachments/DA-23-291A1.pdf""}]","{""kind"": ""Official documents""}" AIBD-0782,2021-04-01,source publication,day,tested,FDA finds that the Gili Pro BioSensor measured two vital signs without contact within its limits for accuracy in a study of 120 adults,"In a study of 120 adults that FDA reviewed, the Gili Pro BioSensor met limits of ±3 bpm against ECG and ±2 breaths per minute against capnography, without touching patients.",*Gili Pro BioSensor* measured two vital signs without contact,limits of ±3 bpm against ECG,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Gili Pro BioSensor (Gili BioSensor System),,ContinUse Biometrics Ltd.,Israel,ContinUse Biometrics (clinical accuracy study in 120 adults against ECG and capnography),Israel,,,Healthcare,,Health,Prediction from data,Patients,3,,physical_health,Hazard reduction,,,,,,,,organizational,2,0,Pre-specified acceptance limits against ECG and capnography,Limits of ±3 bpm for heart rate and ±2 breaths per minute for breathing rate,"Both limits met, with intercepts of 0.51 for heart rate and 0.17 for breathing rate",measured,unknown,,not shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200038.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo Classification Request for Gili Pro BioSensor (Gili BioSensor System) (DEN200038): decision summary. Granted 2021-04-01."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN200038.pdf""}]","{""kind"": ""Official documents""}" @@ -289,7 +289,7 @@ AIBD-1078,2020-11-23,source publication,day,tested,"China's CMDE finds that doct AIBD-0892,2020-11-11,source publication,day,tested,"Japan's transport ministry approved Honda's Traffic Jam Pilot after a safety examination, in the world's first type designation of a Level 3 car",On 11 November 2020 Japan's transport ministry approved the Honda Legend with Traffic Jam Pilot after an examination against the safety standards. The system drives in traffic jams on expressways.,Japan approves *Traffic Jam Pilot* for driving in expressway jams,,,found,Found by a regulator,no comparison,very low,benefit created,false,Traffic Jam Pilot (automated operation device) on the Honda Legend,,"Honda Motor Co., Ltd.",Japan,"Honda Motor Co., Ltd. (factory-fitted on the Legend)",Japan,Japan,Japan,Road vehicle automated driving,,Safety and rescue,Machines that act,Drivers and travelers,11,High income,physical_health,,,,,,,,,personal,2,0,,,,,withdrawn,2021-06-17,not shown,https://www.mlit.go.jp/report/press/jidosha08_hh_003888.html,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Regulatory document"", ""value"": ""Ministry of Land, Infrastructure, Transport and Tourism (Japan), Road Transport Bureau. 世界初!自動運転車(レベル3)の型式指定を行いました (World first: type designation of a Level 3 automated vehicle). Press release, 11 November 2020, section 1."", ""url"": ""https://www.mlit.go.jp/report/press/jidosha08_hh_003888.html""}, {""label"": ""Follow-up"", ""value"": ""Honda told Carscoops in June 2021 that production of the Legend, the car that carries Traffic Jam Pilot, would end for the Japanese market at the end of 2021 as the Sayama plant closed."", ""url"": ""https://www.carscoops.com/2021/06/death-of-jdm-honda-odyssey-wont-impact-us-spec-model/""}]","{""kind"": ""Withdrawals""}" AIBD-1079,2020-10-27,source publication,day,tested,"China's CMDE finds that CoronaryDoc met its targets for triage of stenosis in the maker's trial, judging that it can shorten reading time","In the maker's trial at three institutions that China's CMDE reviewed, CoronaryDoc's triage of coronary stenosis met its target for sensitivity. The same held for specificity. No figures are given.",*CoronaryDoc* met its targets for triage of stenosis in China's review,,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,CoronaryDoc (冠脉CT造影图像血管狭窄辅助分诊软件),Release 2.0 (full version 2.0.0.005),"Yukun (Beijing) Network Technology Co., Ltd. (语坤(北京)网络科技有限公司)",China,Yukun (Beijing) Network Technology (retrospective blinded registration trial at three institutions),China,"Three qualified medical institutions that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers,3,Upper middle income,operational_improvement; physical_health,Enhanced productivity,,,,,,,,organizational,1,0,"Manual reading as the gold standard, with preset targets for sensitivity and specificity",Preset targets (values not given),Sensitivity and specificity reached the targets (no figures given),measured,persisted,2020-11-10,shown,https://www.cmde.org.cn/directory/web/cmde/images/udrC9kNU1OzTsM28zHRqrncz8HVrbio1vq31tXvyO28qOoQ1FaMjAwMDE0M6Opcy5wZGY=.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:冠脉CT造影图像血管狭窄辅助分诊软件 (technical review report for the registration of coronary CT angiography stenosis triage-aid software), acceptance no. CQZ2000143, applicant 语坤(北京)网络科技有限公司. 27 October 2020. Section 五、临床受益-风险评估 (clinical benefit-risk assessment)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/udrC9kNU1OzTsM28zHRqrncz8HVrbio1vq31tXvyO28qOoQ1FaMjAwMDE0M6Opcy5wZGY=.pdf""}]","{""kind"": ""Official documents""}" AIBD-0804,2020-10-02,source publication,day,tested,"FDA found that examiners matched manual microscopy on 645 blood specimens with Scopio's X100, which classifies white cells with a neural network","In FDA's review of the X100 with Full Field PBS, examiners used its neural network to classify white cells. They matched manual microscopy on 645 specimens, with 96.29% agreement.",*X100 with Full Field PBS* matched manual microscopy on 645 smears,"matched manual microscopy on 645 specimens, with 96.29% agreement",,found,Found by a regulator,tested against a standard,very low,aggregate result,false,X100 with Full Field Peripheral Blood Smear Application,,Scopio Labs Ltd.,Israel,Scopio Labs (method comparison against manual microscopy at three sites),Israel,,,Healthcare,,Health,Images and video,Health workers,3,,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,Manual light microscopy of the same 645 specimens,Manual differential (reference),"Agreement 96.29%, sensitivity 87.86% and specificity 97.62% for white-cell differentials",measured,persisted,2022-02-09,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/K201301.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. 510(k) substantial equivalence determination decision summary for X100 with Full Field Peripheral Blood Smear (PBS) Application (K201301). Cleared 2020-10-02."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/K201301.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0984,2020-09-03,source publication,day,intended,A Chicago federal court found that the City's plan to review 1.3 million pages with Relativity Active Learning was reasonable,In September 2020 a US magistrate judge found that Chicago's plan to review 1.3 million pages with Relativity Active Learning was reasonable. The judge found that it was also proportional to the case.,Court backs Chicago's *Active Learning* review of 1.3 million pages,1.3 million pages,,found,Found by a court,,not rated,benefit created,false,Relativity Active Learning (AL),,Relativity (Relativity ODA LLC),United States,City of Chicago (defendant) and its attorneys,United States,"Chicago, Illinois, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Value above replacement,,,,,,,,organizational,1,0,Manual review or keyword searches,,,assumed,unknown,,shown,https://storage.courtlistener.com/recap/gov.uscourts.ilnd.332956/gov.uscourts.ilnd.332956.309.0_1.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Northern District of Illinois (Magistrate Judge Young B. Kim). Livingston v. City of Chicago, No. 16 CV 10156, Memorandum Opinion and Order (Doc. 309), 3 September 2020. CourtListener RECAP copy of the filed order."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.ilnd.332956/gov.uscourts.ilnd.332956.309.0_1.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0984,2020-09-03,source publication,day,intended,A Chicago federal court found that the City's plan to review 1.3 million pages with Relativity Active Learning was reasonable,In September 2020 a US magistrate judge found that Chicago's plan to review 1.3 million pages with Relativity Active Learning was reasonable. The judge found that it was also proportional to the case.,Court backs Chicago's *Active Learning* review of 1.3 million pages,1.3 million pages,,found,Found by a court,,very low,benefit created,false,Relativity Active Learning (AL),,Relativity (Relativity ODA LLC),United States,City of Chicago (defendant) and its attorneys,United States,"Chicago, Illinois, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Value above replacement,,,,,,,,organizational,1,0,Manual review or keyword searches,,,assumed,unknown,,shown,https://storage.courtlistener.com/recap/gov.uscourts.ilnd.332956/gov.uscourts.ilnd.332956.309.0_1.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Northern District of Illinois (Magistrate Judge Young B. Kim). Livingston v. City of Chicago, No. 16 CV 10156, Memorandum Opinion and Order (Doc. 309), 3 September 2020. CourtListener RECAP copy of the filed order."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.ilnd.332956/gov.uscourts.ilnd.332956.309.0_1.pdf""}]","{""kind"": ""Official documents""}" AIBD-1080,2020-07-31,source publication,day,tested,China's CMDE finds that AIDRscreening met its targets for accuracy on stage II or worse diabetic retinopathy in the maker's trials,"In the maker's prospective multicenter trial that China's CMDE reviewed, AIDRscreening detected stage II or worse diabetic retinopathy with accuracy that met its targets. No figures are given.",*AIDRscreening* met its targets for retinopathy in China's review,,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,AIDRscreening (糖尿病视网膜病变眼底图像辅助诊断软件),Release V01 (full version V01.00.00.00),"Shenzhen SiBionics Intelligent Technology Co., Ltd. (深圳硅基智能科技有限公司)",China,Shenzhen SiBionics Intelligent Technology (prospective multicenter and retrospective registration trials),China,"Trial centers that the report does not name, for a registration in China",China,Healthcare,,Health,Images and video,Health workers; Patients,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,"Senior doctors' readings as the gold standard, with preset accuracy targets",Preset targets (values not given),Diagnostic accuracy reached the targets (no figures given),measured,persisted,2020-08-11,shown,https://www.cmde.org.cn/directory/web/cmde/images/zMfE8rKhytPN+MSksqGx5NHbtdfNvMxuKjW+tXvtsI7bz+o6hDUVoxOTAwNjY4o6lzLnBkZg==.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:糖尿病视网膜病变眼底图像辅助诊断软件 (technical review report for the registration of diabetic retinopathy fundus image diagnosis-aid software), acceptance no. CQZ1900668, applicant 深圳硅基智能科技有限公司. 31 July 2020. Section 三、临床评价摘要 (clinical evaluation summary)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/zMfE8rKhytPN+MSksqGx5NHbtdfNvMxuKjW+tXvtsI7bz+o6hDUVoxOTAwNjY4o6lzLnBkZg==.pdf""}]","{""kind"": ""Official documents""}" AIBD-1081,2020-07-31,source publication,day,tested,China's CMDE finds that Airdoc-AIFUNDUS met its targets for accuracy on diabetic retinopathy in the maker's trial,"In the maker's trial that China's CMDE reviewed, Airdoc-AIFUNDUS reached its preset target for sensitivity to diabetic retinopathy. The same held for specificity. The report gives no figures.",*Airdoc-AIFUNDUS* met its targets for retinopathy in China's review,,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Airdoc-AIFUNDUS (糖尿病视网膜病变眼底图像辅助诊断软件),Release V1.0 (full version V1.0.2020.04.02.01),"Shanghai Airdoc Medical Technology Co., Ltd. (上海鹰瞳医疗科技有限公司)",China,Shanghai Airdoc Medical Technology (prospective multicenter registration trial),China,Trial sites in China that the report does not name,China,Healthcare,,Health,Images and video,Health workers; Patients,3,Upper middle income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,"Preset targets for sensitivity and specificity, with senior doctors' reading as the reference standard",Preset targets that the report does not state,Sensitivity and specificity reached the targets,measured,scaled,2026-03-31,shown,https://www.cmde.org.cn/directory/web/cmde/images/zMfE8rKhytPN+MSksqGx5NHbtdfNvMxuKjW+tXvtsI7bz+o6hDUVoxOTAwNjUzo6lzLnBkZg==.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:糖尿病视网膜病变眼底图像辅助诊断软件 (technical review report for the registration of diabetic retinopathy fundus image diagnosis-aid software), acceptance no. CQZ1900653, applicant 上海鹰瞳医疗科技有限公司. 31 July 2020. Section 三、临床评价摘要 (clinical evaluation summary)."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/zMfE8rKhytPN+MSksqGx5NHbtdfNvMxuKjW+tXvtsI7bz+o6hDUVoxOTAwNjUzo6lzLnBkZg==.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Center for Medical Device Evaluation (CMDE), National Medical Products Administration, China. 医疗器械产品注册技术审评报告:糖尿病视网膜病变眼底图像辅助诊断软件 (technical review report for the registration of diabetic retinopathy fundus image diagnosis-aid software), acceptance no. CQZ1900653, applicant 上海鹰瞳医疗科技有限公司. 31 July 2020. Section 综合评价意见 (concluding evaluation), CMDE's own recommendation to approve, with a condition for further study after approval."", ""url"": ""https://www.cmde.org.cn/directory/web/cmde/images/zMfE8rKhytPN+MSksqGx5NHbtdfNvMxuKjW+tXvtsI7bz+o6hDUVoxOTAwNjUzo6lzLnBkZg==.pdf""}]","{""kind"": ""Official documents""}" AIBD-1011,2020-07-01,event,month,mixed,South Korea's Board of Audit and Inspection finds that the AI job recommendation service scores evenly across occupations but drew fewer applications,South Korea's audit board found in December 2025 that the AI job recommender scored about 80% for top and other occupations alike. In 2022 it found lower application rates than with an older system.,*AI job recommendation service* scores evenly but drew fewer applicants,scored about 80% for top and other occupations alike,,found,Found by an auditor,tested against a standard,low,aggregate result,false,"AI 일자리 추천서비스 (AI job recommendation service), called The Work AI on Worknet",,"Korea Employment Information Service (한국고용정보원), for the Ministry of Employment and Labor",South Korea,"Korea Employment Information Service, on the Worknet jobs portal of the Ministry of Employment and Labor",South Korea,South Korea (national),South Korea,Public employment services (job matching),,Public sector,Planning and control,Other workers,8,High income,opportunity_access,Employment access,,,4,3,2,,,personal,2,0,"The older rule-based recommender on Worknet (The Work) in the same periods, from July 2020 to October 2021",15.22% and 18.36% of views led to applications,9.45% and 12.99% of views of AI recommendations led to applications,measured,scaled,2026-10-28,shown,https://www.bai.go.kr/api/files/download?fileId=jj20220412%20141606&fileSn=1,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Board of Audit and Inspection of Korea (감사원). 감사보고서 - 인공지능 대비실태 Ⅰ(신뢰성 확보 분야) - [Performance audit report, AI preparedness I: ensuring reliability]. Report dated December 2025, published 5 January 2026. Chapter 3, section 4, printed p. 54 and footnotes 52 and 53."", ""url"": ""https://www.bai.go.kr/api/files/download?fileId=jj20260303%20152833&fileSn=1""}, {""label"": ""Audit report"", ""value"": ""Board of Audit and Inspection of Korea (감사원). 감사보고서 - 취업알선정보망 구축 및 관리실태 - [Performance audit report, construction and management of the employment placement information network]. Report dated March 2022, released April 2022. Printed pp. 20-24 and [표 13]."", ""url"": ""https://www.bai.go.kr/api/files/download?fileId=jj20220412%20141606&fileSn=1""}]","{""kind"": ""Official documents""}" @@ -316,19 +316,19 @@ AIBD-0806,2017-06-01,event,month,tested,FDA found that the FluChip-8G Influenza AIBD-2303,2017-04-01,event,month,tested,"IQWiG's benefit assessment of fundus photography: in the study with EyeArt as the only reader, photographs found 96.6% of diabetic retinopathy that needs work-up, against 28.0% for ophthalmoscopy","IQWiG drew on a U.S. study in which EyeArt alone read retinal photographs of 521 people with diabetes. It found 96.6% of retinopathy that needs work-up, against 28.0% for dilated eye exams.",*EyeArt* found more diabetic eye disease than dilated eye exams,96.6%,,found,Found by an assessment body,tested against a standard,very low,program,false,EyeArt,,"Eyenuk, Inc.",United States,"10 study centers in the United States (trial), with an assessment for German statutory health care",United States,United States (trial) and Germany (assessment),United States | Germany,Healthcare (diabetic eye screening),,Health,Images and video,Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,Ophthalmologists' examination of the same people with dilated pupils,"Sensitivity 28.0%, specificity 99.6%","Sensitivity 96.6%, specificity 86.7%",measured,unknown,,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC9636573/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}, {""label"": ""Assessment report"", ""value"": ""Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG). Fundusfotografie zur Diagnostik der diabetischen Retinopathie. Abschlussbericht D25-01, dated 9 July 2026 and published in August 2026. doi:10.60584/D25-01. IQWiG is Germany's institute for the assessment of benefits in health care, and the Federal Joint Committee commissioned the report. The quotations are in German."", ""url"": ""https://www.iqwig.de/download/d25-01_fundusfotografie-zur-diagnostik-der-diabetischen-retinopathie_abschlussbericht_v1-0.pdf""}]","{""kind"": ""Official documents""}" AIBD-0792,2017-03-24,source publication,day,tested,FDA finds that radiologists read 3D mammograms 29.2% faster with iCAD's PowerLook Tomo Detection Software at non-inferior accuracy,"In a reader study that FDA reviewed, 20 radiologists read 240 tomosynthesis exams 29.2% faster with PowerLook Tomo Detection Software. Their AUC was non-inferior.",*PowerLook Tomo* cut radiologists' reading time on 3D mammograms,29.2% faster,,found,Found by a regulator,controlled comparison,very low,aggregate result,false,iCAD PowerLook Tomo Detection Software,,"iCAD, Inc.",United States,iCAD (reader study with 20 radiologists on 240 GE tomosynthesis cases),United States,"San Antonio, Texas, United States",United States,Healthcare,,Health,Images and video,Health workers,3,High income,operational_improvement,Enhanced productivity,,2,1,,,1,,organizational,2,0,The same radiologists reading the same cases without the software,AUC 0.841,"AUC 0.850, with reading time 29.2% shorter",measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf16/P160009B.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. Summary of Safety and Effectiveness Data (SSED), P160009: PowerLook Tomo Detection Software. Approved 2017-03-24."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf16/P160009B.pdf""}]","{""kind"": ""Official documents""}" AIBD-0928,2017-03-01,event,month,tested,France's HAS rated the DBLG1 closed-loop insulin system a moderate improvement over a pump used with a sensor after a 12-week trial,HAS found that people using the DBLG1 closed loop spent 68.5% of the time in the target glucose range against 59.4% on pump and sensor alone. It rated the gain as moderate.,*DBLG1* closed loop kept glucose in range longer in a French trial,68.5% of the time in the target glucose range against 59.4%,,found,Found by an assessment body,randomized trial,moderate,aggregate result,false,DBLG1 System,DBLG1 algorithm v2017.04.20 in the SP7 trial (listed system: software version 1.3.0 of 16 May 2019),Diabeloop,France,12 French diabetes centers in the SP7 randomized crossover trial,France,France (12 trial centers),France,Healthcare (type 1 diabetes),,Health,Prediction from data; Machines that act,Patients,3,High income,physical_health,,,,3,,,5,,personal,1,0,Open-loop systems in which an insulin pump and a glucose sensor work independently,"59,4 (± 1,3) % of time in 70-180 mg/dL with pump and sensor in the SP7 trial","68,5 (± 1,2) % with DBLG1, rated amélioration modérée du service attendu (ASA III)",measured,scaled,2024-12-17,shown,https://www.has-sante.fr/upload/docs/evamed/CNEDIMTS-5998_DBLG1%205998_occultation.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Haute Autorité de santé, Commission nationale d'évaluation des dispositifs médicaux et des technologies de santé (CNEDiMTS). Avis de la CNEDiMTS du 28 janvier 2020, DBLG1, système de boucle semi-fermée dédié à la gestion automatisée du diabète de type 1 (dossier 5998, version publiée après occultation). Section 06.2, Niveau(x) d'ASA."", ""url"": ""https://www.has-sante.fr/upload/docs/evamed/CNEDIMTS-5998_DBLG1%205998_occultation.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0832,2016-12-02,source publication,day,intended,"The Supreme Court of Victoria ordered technology assisted review for about 1.4 million documents that would take over 23,000 hours to review by hand","In December 2016 the Supreme Court of Victoria ordered technology assisted review for about 1.4 million documents. It said that manual review would take a junior solicitor over 23,000 hours.",Victorian court orders *TAR* for 1.4 million pipeline documents,"over 23,000 hours",,found,Found by a court,,not rated,benefit created,false,Predictive coding technology (TAR or Technology Assisted Review),,Not named in the judgment,,"McConnell Dowell Constructors (Aust) Pty Ltd (plaintiff), with the agreement of all parties",Australia,"Melbourne, Australia",Australia,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,"Manual review of 1,400,000 documents at one minute each","over 23,000 hours, equating to over 583 working weeks",,assumed,persisted,,shown,https://web.archive.org/web/20251018042102if_/https://www.austlii.edu.au/cgi-bin/viewdoc/au/cases/vic/VSC/2016/734.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Supreme Court of Victoria, Commercial Court, Technology, Engineering and Construction List (Vickery J). McConnell Dowell Constructors (Aust) Pty Ltd v Santam Ltd & Ors (No 1) [2016] VSC 734, 2 December 2016. AustLII copy, read through the Internet Archive."", ""url"": ""https://web.archive.org/web/20251018042102if_/https://www.austlii.edu.au/cgi-bin/viewdoc/au/cases/vic/VSC/2016/734.html""}]","{""kind"": ""Official documents""}" +AIBD-0832,2016-12-02,source publication,day,intended,"The Supreme Court of Victoria ordered technology assisted review for about 1.4 million documents that would take over 23,000 hours to review by hand","In December 2016 the Supreme Court of Victoria ordered technology assisted review for about 1.4 million documents. It said that manual review would take a junior solicitor over 23,000 hours.",Victorian court orders *TAR* for 1.4 million pipeline documents,"over 23,000 hours",,found,Found by a court,,very low,benefit created,false,Predictive coding technology (TAR or Technology Assisted Review),,Not named in the judgment,,"McConnell Dowell Constructors (Aust) Pty Ltd (plaintiff), with the agreement of all parties",Australia,"Melbourne, Australia",Australia,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,"Manual review of 1,400,000 documents at one minute each","over 23,000 hours, equating to over 583 working weeks",,assumed,persisted,,shown,https://web.archive.org/web/20251018042102if_/https://www.austlii.edu.au/cgi-bin/viewdoc/au/cases/vic/VSC/2016/734.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Supreme Court of Victoria, Commercial Court, Technology, Engineering and Construction List (Vickery J). McConnell Dowell Constructors (Aust) Pty Ltd v Santam Ltd & Ors (No 1) [2016] VSC 734, 2 December 2016. AustLII copy, read through the Internet Archive."", ""url"": ""https://web.archive.org/web/20251018042102if_/https://www.austlii.edu.au/cgi-bin/viewdoc/au/cases/vic/VSC/2016/734.html""}]","{""kind"": ""Official documents""}" AIBD-0793,2016-11-09,source publication,day,tested,"FDA finds that radiologists read automated breast ultrasound a third faster with QView's QVCAD, with non-inferior cancer detection","In a reader study that FDA reviewed, 18 radiologists took 144 seconds per ultrasound case with QVCAD against 215 seconds without it, with non-inferior AUC.",*QVCAD* cut reading time for breast ultrasound by a third,a 71-second or 33% reduction,,found,Found by a regulator,controlled comparison,very low,aggregate result,false,QVCAD System,,"QView Medical, Inc.",United States,QView Medical (reader study with 18 radiologists on 185 automated breast ultrasound cases),United States,United States,United States,Healthcare,,Health,Images and video,Health workers,3,High income,operational_improvement,Enhanced productivity,,2,1,,,1,,organizational,2,0,The same radiologists reading the same cases without QVCAD,215 seconds per case,144 seconds per case,measured,scaled,2017-12-21,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf15/P150043B.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. Summary of Safety and Effectiveness Data (SSED), P150043: QVCAD System. Approved 2016-11-09."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf15/P150043B.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0831,2016-05-17,source publication,day,intended,"The High Court in London ordered predictive coding over a petitioner's objection, finding that it would cost about £132,000 against at least £250,000","In May 2016 Mr Registrar Jones ordered disclosure by predictive coding over the petitioner's objection. He found that it would cost about £132,000, against at least £250,000 for a keyword search.","Court orders *predictive coding* at £132,000 over a keyword search","about £132,000, against at least £250,000",,found,Found by a court,,not rated,benefit created,false,Predictive coding,,Not named in the judgment,,BCA Trading Ltd and the other respondents,United Kingdom,"London, United Kingdom",United Kingdom,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,A keyword search approach to disclosure,"at least £250,000 (and indeed I believe that figure may even reach £338,000 on a worst case scenario)","in the region of £132,000",assumed,persisted,2018-08-28,shown,https://web.archive.org/web/20251018042102if_/https://www.austlii.edu.au/cgi-bin/viewdoc/au/cases/vic/VSC/2016/734.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""High Court of Justice, Chancery Division, Companies Court (Mr Registrar Jones). Brown v BCA Trading Ltd & Ors [2016] EWHC 1464 (Ch), approved judgment of 17 May 2016. Find Case Law copy."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/1464""}]","{""kind"": ""Official documents""}" -AIBD-0830,2016-02-16,source publication,day,intended,"The High Court in London approved predictive coding software for over 3 million documents, finding no evidence that it gives less accurate disclosure",In February 2016 Master Matthews approved predictive coding for over 3 million documents. He found no evidence that it gives less accurate disclosure. He judged it far cheaper than manual review.,English court approves *predictive coding* for 3 million documents,over 3 million documents,,found,Found by a court,,not rated,benefit created,false,Predictive coding software,,Not named in the judgment,,"MWB Business Exchange Ltd (second claimant), with the agreement of all parties",United Kingdom,"London, United Kingdom",United Kingdom,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,A full manual review of over 3 million documents,several million pounds at least,"£181,988 plus monthly hosting costs of £15,717, to £469,049 plus monthly hosting costs of £20,820",assumed,unknown,,shown,https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/256,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""High Court of Justice, Chancery Division (Master Matthews). Pyrrho Investments Ltd & Anor v MWB Property Ltd & Ors [2016] EWHC 256 (Ch), approved judgment of 16 February 2016. Find Case Law copy, a scan of the approved judgment."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/256""}, {""label"": ""Court document"", ""value"": ""High Court of Justice, Chancery Division, Companies Court (Mr Registrar Jones). Brown v BCA Trading Ltd & Ors [2016] EWHC 1464 (Ch), 17 May 2016, para. 10, quoting Pyrrho para. 33 with its cost estimates."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/1464""}]","{""kind"": ""Official documents""}" -AIBD-0829,2015-03-03,source publication,day,intended,"The High Court of Ireland approved technology assisted review for 680,809 documents, finding predictive coding at least as accurate as manual review","In March 2015 the High Court of Ireland approved predictive coding for 680,809 documents over the defendants' objection. It found that the method was at least as accurate as manual review.","Irish court backs *predictive coding* for 680,809 bank documents","less than 10% of the 680,809 documents",,found,Found by a court,,not rated,benefit created,false,Technology assisted review using predictive coding,,"Not named in the judgment (the bank's expert took his target f-measure from Symantec, the provider of the Clearwell system)",,"Irish Bank Resolution Corporation and its solicitors, in discovery",Ireland,"Dublin, Ireland",Ireland,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,"Manual (linear) review of the 680,809 documents","9 months at a cost of €2 m by the bank's solicitor's estimate, or 113 days at €220,000 by the defendants' expert's","less than 10% of the 680,809 documents would need to be manually reviewed",assumed,persisted,2016-03-01,shown,https://web.archive.org/web/20240428025752if_/https://www.bailii.org/ie/cases/IEHC/2015/H175.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""High Court of Ireland (Fullam J.). Irish Bank Resolution Corporation Ltd & ors v Quinn & ors [2015] IEHC 175, judgment of 3 March 2015. BAILII copy, read through the Internet Archive."", ""url"": ""https://web.archive.org/web/20240428025752if_/https://www.bailii.org/ie/cases/IEHC/2015/H175.html""}]","{""kind"": ""Official documents""}" -AIBD-0981,2015-03-02,source publication,day,intended,"A New York federal court approved Rio Tinto and Vale's agreed TAR protocol, stressing that TAR must face no higher standard than manual review",In March 2015 a US magistrate judge approved Rio Tinto and Vale's protocol for predictive coding. He wrote that TAR must face no higher standard than manual review. He wrote the same of keywords.,Federal court approves a *TAR* protocol agreed by Rio Tinto and Vale,,,found,Found by a court,,not rated,benefit created,false,"Relativity Assisted Review (Rio Tinto) and Deloitte's Dynamic Review (Vale), under a stipulated predictive coding protocol",,"Relativity (Relativity Assisted Review) and Deloitte (Dynamic Review), as the parties' letter attached to the opinion names the tools",United States (Relativity); not verified for Deloitte,"Rio Tinto plc (plaintiff), with its vendor Precision Discovery, and Vale S.A. (defendant)",United Kingdom and Brazil,"New York, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,Keyword searches or manual review,,,assumed,persisted,2015-07-15,shown,https://static.case.law/frd/306/html/0125-01.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Southern District of New York (Magistrate Judge Andrew J. Peck). Rio Tinto PLC v. Vale S.A., 306 F.R.D. 125 (S.D.N.Y. 2015), Opinion and Order of 2 March 2015, with the parties' letter and stipulated protocol attached as an appendix. Caselaw Access Project copy of the Federal Rules Decisions report."", ""url"": ""https://static.case.law/frd/306/html/0125-01.html""}]","{""kind"": ""Official documents""}" +AIBD-0831,2016-05-17,source publication,day,intended,"The High Court in London ordered predictive coding over a petitioner's objection, finding that it would cost about £132,000 against at least £250,000","In May 2016 Mr Registrar Jones ordered disclosure by predictive coding over the petitioner's objection. He found that it would cost about £132,000, against at least £250,000 for a keyword search.","Court orders *predictive coding* at £132,000 over a keyword search","about £132,000, against at least £250,000",,found,Found by a court,,very low,benefit created,false,Predictive coding,,Not named in the judgment,,BCA Trading Ltd and the other respondents,United Kingdom,"London, United Kingdom",United Kingdom,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,A keyword search approach to disclosure,"at least £250,000 (and indeed I believe that figure may even reach £338,000 on a worst case scenario)","in the region of £132,000",assumed,persisted,2018-08-28,shown,https://web.archive.org/web/20251018042102if_/https://www.austlii.edu.au/cgi-bin/viewdoc/au/cases/vic/VSC/2016/734.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""High Court of Justice, Chancery Division, Companies Court (Mr Registrar Jones). Brown v BCA Trading Ltd & Ors [2016] EWHC 1464 (Ch), approved judgment of 17 May 2016. Find Case Law copy."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/1464""}]","{""kind"": ""Official documents""}" +AIBD-0830,2016-02-16,source publication,day,intended,"The High Court in London approved predictive coding software for over 3 million documents, finding no evidence that it gives less accurate disclosure",In February 2016 Master Matthews approved predictive coding for over 3 million documents. He found no evidence that it gives less accurate disclosure. He judged it far cheaper than manual review.,English court approves *predictive coding* for 3 million documents,over 3 million documents,,found,Found by a court,,very low,benefit created,false,Predictive coding software,,Not named in the judgment,,"MWB Business Exchange Ltd (second claimant), with the agreement of all parties",United Kingdom,"London, United Kingdom",United Kingdom,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,A full manual review of over 3 million documents,several million pounds at least,"£181,988 plus monthly hosting costs of £15,717, to £469,049 plus monthly hosting costs of £20,820",assumed,unknown,,shown,https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/256,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""High Court of Justice, Chancery Division (Master Matthews). Pyrrho Investments Ltd & Anor v MWB Property Ltd & Ors [2016] EWHC 256 (Ch), approved judgment of 16 February 2016. Find Case Law copy, a scan of the approved judgment."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/256""}, {""label"": ""Court document"", ""value"": ""High Court of Justice, Chancery Division, Companies Court (Mr Registrar Jones). Brown v BCA Trading Ltd & Ors [2016] EWHC 1464 (Ch), 17 May 2016, para. 10, quoting Pyrrho para. 33 with its cost estimates."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/1464""}]","{""kind"": ""Official documents""}" +AIBD-0829,2015-03-03,source publication,day,intended,"The High Court of Ireland approved technology assisted review for 680,809 documents, finding predictive coding at least as accurate as manual review","In March 2015 the High Court of Ireland approved predictive coding for 680,809 documents over the defendants' objection. It found that the method was at least as accurate as manual review.","Irish court backs *predictive coding* for 680,809 bank documents","less than 10% of the 680,809 documents",,found,Found by a court,,very low,benefit created,false,Technology assisted review using predictive coding,,"Not named in the judgment (the bank's expert took his target f-measure from Symantec, the provider of the Clearwell system)",,"Irish Bank Resolution Corporation and its solicitors, in discovery",Ireland,"Dublin, Ireland",Ireland,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,"Manual (linear) review of the 680,809 documents","9 months at a cost of €2 m by the bank's solicitor's estimate, or 113 days at €220,000 by the defendants' expert's","less than 10% of the 680,809 documents would need to be manually reviewed",assumed,persisted,2016-03-01,shown,https://web.archive.org/web/20240428025752if_/https://www.bailii.org/ie/cases/IEHC/2015/H175.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""High Court of Ireland (Fullam J.). Irish Bank Resolution Corporation Ltd & ors v Quinn & ors [2015] IEHC 175, judgment of 3 March 2015. BAILII copy, read through the Internet Archive."", ""url"": ""https://web.archive.org/web/20240428025752if_/https://www.bailii.org/ie/cases/IEHC/2015/H175.html""}]","{""kind"": ""Official documents""}" +AIBD-0981,2015-03-02,source publication,day,intended,"A New York federal court approved Rio Tinto and Vale's agreed TAR protocol, stressing that TAR must face no higher standard than manual review",In March 2015 a US magistrate judge approved Rio Tinto and Vale's protocol for predictive coding. He wrote that TAR must face no higher standard than manual review. He wrote the same of keywords.,Federal court approves a *TAR* protocol agreed by Rio Tinto and Vale,,,found,Found by a court,,very low,benefit created,false,"Relativity Assisted Review (Rio Tinto) and Deloitte's Dynamic Review (Vale), under a stipulated predictive coding protocol",,"Relativity (Relativity Assisted Review) and Deloitte (Dynamic Review), as the parties' letter attached to the opinion names the tools",United States (Relativity); not verified for Deloitte,"Rio Tinto plc (plaintiff), with its vendor Precision Discovery, and Vale S.A. (defendant)",United Kingdom and Brazil,"New York, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,Keyword searches or manual review,,,assumed,persisted,2015-07-15,shown,https://static.case.law/frd/306/html/0125-01.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Southern District of New York (Magistrate Judge Andrew J. Peck). Rio Tinto PLC v. Vale S.A., 306 F.R.D. 125 (S.D.N.Y. 2015), Opinion and Order of 2 March 2015, with the parties' letter and stipulated protocol attached as an appendix. Caselaw Access Project copy of the Federal Rules Decisions report."", ""url"": ""https://static.case.law/frd/306/html/0125-01.html""}]","{""kind"": ""Official documents""}" AIBD-0786,2014-11-17,source publication,day,tested,FDA finds that the BrainScope Ahead 100 missed its endpoints for accuracy but ruled out brain injury on CT with an 89.5% negative predictive value,"In FDA's review of the BrainScope Ahead 100, the EEG test missed its endpoints for accuracy in 552 patients with head injury. Its negative predictive value for injury on CT was 89.5%.",*Ahead 100* EEG test missed its goals but helped rule out injury on CT,negative predictive value for injury on CT was 89.5%,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,BrainScope Ahead 100 (Models CV-100 and M-100),,"BrainScope Company, Inc.",United States,BrainScope (B-AHEAD II study at 11 US emergency department sites),United States,"Eleven emergency department sites, United States",United States,Healthcare,,Health,Prediction from data,Patients; Health workers,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,Pre-specified endpoints for the lower confidence limits of sensitivity and specificity,Lower limits above 78% for sensitivity and 50% for specificity,"Sensitivity 78.5% and specificity 48.6%, with the endpoints missed",measured,persisted,2015-05-15,shown,https://www.brainscope.com/sciencetechnology,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo Classification Request for BrainScope Ahead 100, Models CV-100 and M-100 (DEN140025): decision summary. Granted 2014-11-17."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN140025.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0833,2014-09-17,source publication,day,intended,"The US Tax Court let two taxpayers answer a discovery request from the IRS with predictive coding, crediting an estimate of $80,000 to $85,000 for review","In September 2014 the US Tax Court allowed predictive coding over the IRS's objection. It credited an estimate of 200,000 to 400,000 documents to review at $80,000 to $85,000.",Tax Court lets taxpayers use *predictive coding* though the IRS objects,"200,000 to 400,000 documents to review at $80,000 to $85,000",,found,Found by a court,,not rated,benefit created,false,Predictive coding,,Not named in the opinion,,"Dynamo Holdings Limited Partnership and Beekman Vista, Inc. (petitioners)",United States,United States,United States,Legal services / e-discovery,,Science,Text and language,,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,"Review under the Commissioner's approach, a privilege search of all restored data","3.5 million to 7 million documents at a cost of $500,000 to $550,000","200,000 to 400,000 documents at a cost of $80,000 to $85,000",assumed,litigated,2016-07-13,shown,https://static.case.law/tc/143/html/0183-01.html,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""United States Tax Court (Judge Ronald L. Buch). Dynamo Holdings Ltd. Partnership v. Commissioner, 143 T.C. 183 (2014), opinion of 17 September 2014. Caselaw Access Project copy of the Tax Court report."", ""url"": ""https://static.case.law/tc/143/html/0183-01.html""}, {""label"": ""Follow-up"", ""value"": ""The review ran, and the Commissioner moved to compel more documents. On 13 July 2016 the court denied the motion and held that the response was complete."", ""url"": ""https://www.ediscoverylaw.com/wp-content/uploads/2016/08/Dynamo-II-Order.pdf""}]","{""kind"": ""Legal proceedings""}" +AIBD-0833,2014-09-17,source publication,day,intended,"The US Tax Court let two taxpayers answer a discovery request from the IRS with predictive coding, crediting an estimate of $80,000 to $85,000 for review","In September 2014 the US Tax Court allowed predictive coding over the IRS's objection. It credited an estimate of 200,000 to 400,000 documents to review at $80,000 to $85,000.",Tax Court lets taxpayers use *predictive coding* though the IRS objects,"200,000 to 400,000 documents to review at $80,000 to $85,000",,found,Found by a court,,very low,benefit created,false,Predictive coding,,Not named in the opinion,,"Dynamo Holdings Limited Partnership and Beekman Vista, Inc. (petitioners)",United States,United States,United States,Legal services / e-discovery,,Science,Text and language,,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,"Review under the Commissioner's approach, a privilege search of all restored data","3.5 million to 7 million documents at a cost of $500,000 to $550,000","200,000 to 400,000 documents at a cost of $80,000 to $85,000",assumed,litigated,2016-07-13,shown,https://static.case.law/tc/143/html/0183-01.html,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""United States Tax Court (Judge Ronald L. Buch). Dynamo Holdings Ltd. Partnership v. Commissioner, 143 T.C. 183 (2014), opinion of 17 September 2014. Caselaw Access Project copy of the Tax Court report."", ""url"": ""https://static.case.law/tc/143/html/0183-01.html""}, {""label"": ""Follow-up"", ""value"": ""The review ran, and the Commissioner moved to compel more documents. On 13 July 2016 the court denied the motion and held that the response was complete."", ""url"": ""https://www.ediscoverylaw.com/wp-content/uploads/2016/08/Dynamo-II-Order.pdf""}]","{""kind"": ""Legal proceedings""}" AIBD-0812,2014-08-11,source publication,day,tested,"FDA found that Cologuard detected 92.3% of colorectal cancers against 73.8% for a fecal immunochemical test, at lower specificity","In FDA's review of Cologuard, the stool DNA test detected 92.3% of colorectal cancers against 73.8% for FIT in 9,989 adults. Its specificity was 86.6% against 94.9%.",*Cologuard* caught more colorectal cancers than a standard stool test,detected 92.3% of colorectal cancers against 73.8% for FIT,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Cologuard,,Exact Sciences Corporation,United States,Exact Sciences (DeeP-C study at 90 sites in the United States and Canada),United States,90 sites in the United States and Canada,United States | Canada,Healthcare,,Health,Prediction from data,Patients,3,High income,physical_health,Medical diagnosis,,,,2,,,,organizational,2,0,A commercial fecal immunochemical test (OC FIT-CHEK) in the same adults,"Sensitivity 73.8% for colorectal cancer, specificity 94.9%","Sensitivity 92.3%, specificity 86.6%",measured,scaled,2024-10-04,not shown,https://www.accessdata.fda.gov/cdrh_docs/pdf13/P130017B.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. Summary of Safety and Effectiveness Data for Cologuard (P130017). Approved 2014-08-11."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf13/P130017B.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0983,2014-07-22,source publication,day,intended,"A Tennessee federal court let Bridgestone review over two million documents with predictive coding over IBM's objection, citing efficient discovery","In July 2014 a US magistrate judge let Bridgestone review over two million documents with predictive coding, over IBM's objection. He cited Rule 26's call for efficient discovery.",Court lets Bridgestone use *predictive coding* over IBM's objection,over two million documents,,found,Found by a court,,not rated,benefit created,false,Predictive coding,,Not named in the order,,"Bridgestone Americas, Inc. (plaintiff), with its attorneys, experts and the companies doing the work",United States,"Nashville, Tennessee, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,,,,,challenged,2015-02-05,not shown,https://storage.courtlistener.com/recap/gov.uscourts.tnmd.57186/gov.uscourts.tnmd.57186.89.0.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Middle District of Tennessee (Magistrate Judge Joe B. Brown). Bridgestone Americas, Inc. v. International Business Machines Corp., No. 3:13-1196, Order (Doc. 89), 22 July 2014. CourtListener RECAP copy of the filed order."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.tnmd.57186/gov.uscourts.tnmd.57186.89.0.pdf""}, {""label"": ""Follow-up"", ""value"": ""In January 2015 the parties told the court that some seed set documents listed as nonresponsive were in fact responsive, and the magistrate judge reminded both sides that he expected full transparency. A later opinion in another case recounts this."", ""url"": ""https://static.case.law/frd/306/html/0125-01.html""}]","{""kind"": ""Challenges""}" +AIBD-0983,2014-07-22,source publication,day,intended,"A Tennessee federal court let Bridgestone review over two million documents with predictive coding over IBM's objection, citing efficient discovery","In July 2014 a US magistrate judge let Bridgestone review over two million documents with predictive coding, over IBM's objection. He cited Rule 26's call for efficient discovery.",Court lets Bridgestone use *predictive coding* over IBM's objection,over two million documents,,found,Found by a court,,very low,benefit created,false,Predictive coding,,Not named in the order,,"Bridgestone Americas, Inc. (plaintiff), with its attorneys, experts and the companies doing the work",United States,"Nashville, Tennessee, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,,,,,challenged,2015-02-05,not shown,https://storage.courtlistener.com/recap/gov.uscourts.tnmd.57186/gov.uscourts.tnmd.57186.89.0.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Middle District of Tennessee (Magistrate Judge Joe B. Brown). Bridgestone Americas, Inc. v. International Business Machines Corp., No. 3:13-1196, Order (Doc. 89), 22 July 2014. CourtListener RECAP copy of the filed order."", ""url"": ""https://storage.courtlistener.com/recap/gov.uscourts.tnmd.57186/gov.uscourts.tnmd.57186.89.0.pdf""}, {""label"": ""Follow-up"", ""value"": ""In January 2015 the parties told the court that some seed set documents listed as nonresponsive were in fact responsive, and the magistrate judge reminded both sides that he expected full transparency. A later opinion in another case recounts this."", ""url"": ""https://static.case.law/frd/306/html/0125-01.html""}]","{""kind"": ""Challenges""}" AIBD-0787,2013-08-21,source publication,day,tested,FDA finds that VITEK MS correctly identified 87.5% of 7068 isolates from patient cultures with a machine-learned classifier of spectra,"In a study in five US clinical laboratories that FDA reviewed, VITEK MS gave a correct single identification for 87.5% (6184/7068) of isolates from patient cultures.",*VITEK MS* identified microbes from patient cultures in five labs,87.5% (6184/7068),,found,Found by a regulator,no comparison,very low,aggregate result,false,VITEK MS (MALDI-TOF mass spectrometry with the Advanced Spectrum Classifier),VITEK MS V2.0 Knowledge Base,"bioMérieux, Inc.",United States,Five US clinical microbiology laboratories in the prospective clinical study,United States,"Five clinical microbiology laboratories, United States",United States,Healthcare,,Health,Prediction from data,Health workers,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,1,0,,,,,scaled,2017-07-22,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/K124067.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. Evaluation of Automatic Class III Designation (De Novo) for VITEK MS (K124067 / DEN130013): decision summary. Granted 2013-08-21."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/K124067.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0988,2012-04-23,source publication,day,intended,"The Circuit Court for Loudoun County, Virginia, let the Landow defendants use predictive coding in the Dulles Jet Center cases over objections",In April 2012 a Virginia state court let defendants use predictive coding over objections. The suits concerned a hangar collapse at Dulles Jet Center. The order gives no reasons.,Virginia court lets defendants use *predictive coding* over objections,,,found,Found by a court,,not rated,benefit created,false,Predictive coding,,Not named in the order or in the defendants' memorandum,,"Landow Aviation Limited Partnership, Landow Aviation I, Inc. and Landow & Company Builders, Inc. (defendants)",United States,"Loudoun County, Virginia, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,,,,,persisted,2013-01-21,shown,https://www.ediscoverylaw.com/wp-content/uploads/2013/11/MemoSupportPredictiveCoding.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Circuit Court for Loudoun County, Virginia (Judge James H. Chamblin). Global Aerospace Inc., et al. v. Landow Aviation, L.P. d/b/a Dulles Jet Center, et al., Consolidated Case No. CL 61040, Order Approving the Use of Predictive Coding for Discovery, entered 23 April 2012. Copy of the signed order on the eDiscovery Law blog of K&L Gates."", ""url"": ""https://www.ediscoverylaw.com/wp-content/uploads/2013/11/Order-Approving-Predictive-Coding-.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0828,2012-02-24,source publication,day,intended,"A New York federal court approved predictive coding with Axcelerate for over three million emails, finding it superior to manual review",In February 2012 a US magistrate judge approved predictive coding with Recommind's Axcelerate for over three million emails. He found that it beat manual review.,Federal court approves *Axcelerate* for three million emails,over three million emails,,found,Found by a court,,not rated,benefit created,false,Axcelerate software by Recommind (predictive coding),,Recommind,United States,"MSL Group (defendant), with its counsel Jackson Lewis LLP",United States,"New York, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,Linear manual review or keyword searches,,,assumed,litigated,,shown,https://static.case.law/frd/287/html/0182-01.html,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Southern District of New York (Magistrate Judge Andrew J. Peck). Moore v. Publicis Groupe & MSL Group, 287 F.R.D. 182 (S.D.N.Y. 2012), Opinion and Order, 24 February 2012. Caselaw Access Project copy of the Federal Rules Decisions report."", ""url"": ""https://static.case.law/frd/287/html/0182-01.html""}, {""label"": ""Follow-up"", ""value"": ""The plaintiffs objected, and District Judge Andrew L. Carter Jr. adopted the rulings in April 2012, as Master Matthews recounts in a later English judgment."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/256""}]","{""kind"": ""Legal proceedings""}" +AIBD-0988,2012-04-23,source publication,day,intended,"The Circuit Court for Loudoun County, Virginia, let the Landow defendants use predictive coding in the Dulles Jet Center cases over objections",In April 2012 a Virginia state court let defendants use predictive coding over objections. The suits concerned a hangar collapse at Dulles Jet Center. The order gives no reasons.,Virginia court lets defendants use *predictive coding* over objections,,,found,Found by a court,,very low,benefit created,false,Predictive coding,,Not named in the order or in the defendants' memorandum,,"Landow Aviation Limited Partnership, Landow Aviation I, Inc. and Landow & Company Builders, Inc. (defendants)",United States,"Loudoun County, Virginia, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,,,,,persisted,2013-01-21,shown,https://www.ediscoverylaw.com/wp-content/uploads/2013/11/MemoSupportPredictiveCoding.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Circuit Court for Loudoun County, Virginia (Judge James H. Chamblin). Global Aerospace Inc., et al. v. Landow Aviation, L.P. d/b/a Dulles Jet Center, et al., Consolidated Case No. CL 61040, Order Approving the Use of Predictive Coding for Discovery, entered 23 April 2012. Copy of the signed order on the eDiscovery Law blog of K&L Gates."", ""url"": ""https://www.ediscoverylaw.com/wp-content/uploads/2013/11/Order-Approving-Predictive-Coding-.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0828,2012-02-24,source publication,day,intended,"A New York federal court approved predictive coding with Axcelerate for over three million emails, finding it superior to manual review",In February 2012 a US magistrate judge approved predictive coding with Recommind's Axcelerate for over three million emails. He found that it beat manual review.,Federal court approves *Axcelerate* for three million emails,over three million emails,,found,Found by a court,,very low,benefit created,false,Axcelerate software by Recommind (predictive coding),,Recommind,United States,"MSL Group (defendant), with its counsel Jackson Lewis LLP",United States,"New York, United States",United States,Legal services / e-discovery,,Science,Text and language,Other workers,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,1,0,Linear manual review or keyword searches,,,assumed,litigated,,shown,https://static.case.law/frd/287/html/0182-01.html,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""United States District Court for the Southern District of New York (Magistrate Judge Andrew J. Peck). Moore v. Publicis Groupe & MSL Group, 287 F.R.D. 182 (S.D.N.Y. 2012), Opinion and Order, 24 February 2012. Caselaw Access Project copy of the Federal Rules Decisions report."", ""url"": ""https://static.case.law/frd/287/html/0182-01.html""}, {""label"": ""Follow-up"", ""value"": ""The plaintiffs objected, and District Judge Andrew L. Carter Jr. adopted the rulings in April 2012, as Master Matthews recounts in a later English judgment."", ""url"": ""https://caselaw.nationalarchives.gov.uk/ewhc/ch/2016/256""}]","{""kind"": ""Legal proceedings""}" AIBD-0910,2010-07-01,event,month,tested,EPA approves trained Pavilion8 models in place of hardware analyzers for NOx on seven turbines in Arkansas after they passed its tests of accuracy,"EPA found that trained Pavilion8 models matched hardware analyzers for NOx on seven turbines in Arkansas, with relative accuracy of 1.38% to 3.15%. It approved them in their place.",EPA lets *Pavilion8 PEMS* stand in for NOx analyzers in Arkansas,relative accuracy of 1.38% to 3.15%,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Pavilion8 Software PEMS (predictive emission monitoring systems) on Units G1 through G7,,"Rockwell Automation (maker of Pavilion8, which the letter names only as a product)",United States,"Arkansas Electric Cooperative Corporation (AECC), Harry L. Oswald Generating Station",United States,"Pulaski County, Arkansas, United States",United States,Air emissions monitoring (power generation),,Environment,Prediction from data,,,High income,operational_improvement,,,1,1,,,1,,organizational,1,0,Certified hardware NOx analyzers (CEMS) on the same turbines,,"Relative accuracy of 1.38% to 3.15% across the seven units, with all three statistical tests passed",measured,persisted,2015-03-16,shown,https://www.epa.gov/sites/default/files/2015-05/documents/harry_l._oswald_generating_station_0.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Environmental Protection Agency, Clean Air Markets Division. Letter to the Arkansas Electric Cooperative Corporation approving alternative monitoring systems (PEMS) for Units G1 through G7 at the Harry L. Oswald Generating Station (ORISPL 55221), signed by Reid Harvey, Director, 21 November 2012."", ""url"": ""https://www.epa.gov/sites/default/files/2015-05/documents/harry_l._oswald_generating_station_0.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Environmental Protection Agency, Clean Air Markets Division. Letter to the Arkansas Electric Cooperative Corporation, 21 November 2012. Background."", ""url"": ""https://www.epa.gov/sites/default/files/2015-05/documents/harry_l._oswald_generating_station_0.pdf""}]","{""kind"": ""Official documents""}" AIBD-0905,2010-05-12,source publication,day,tested,FDA found that white cell differentials from Medica's EasyCell correlated 0.93 to 0.99 with manual microscopy after a neural network preclassified the cells,"In FDA's review, technologists reviewed EasyCell's white cell differentials on 304 specimens. These correlated with manual microscopy at 0.93 to 0.99, including 0.99 for neutrophils.",*EasyCell*'s white cell counts track manual microscopy in FDA's review,correlation coefficients of 0.99 for neutrophils,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,EasyCell Cell Locator (EasyCell),,Medica Corporation,United States,Medica Corporation (method comparison study at three sites),United States,Three study sites (not named in the decision summary),,Healthcare (laboratory hematology),,Health,Images and video,Health workers,3,,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,Manual WBC differential by microscopy on the same specimens,Manual 200-cell differential (reference),"Correlation coefficient 0.99 for neutrophils, 0.98 for lymphocytes, 0.97 for eosinophils and 0.93 for monocytes",measured,persisted,2023-04-24,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/K092116.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. 510(k) substantial equivalence determination decision summary, k092116: EasyCell Cell Locator (Medica Corporation). Decision date 12 May 2010."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/K092116.pdf""}, {""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. 510(k) substantial equivalence determination decision summary, k092116: EasyCell Cell Locator (Medica Corporation). Principles of operation."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/K092116.pdf""}]","{""kind"": ""Official documents""}" AIBD-0808,2008-07-30,source publication,day,tested,FDA found that the machine-learned Pathwork Tissue of Origin Test agreed with the known diagnosis for most of 545 tumor specimens,"In FDA's review of the Pathwork Tissue of Origin Test, results agreed with the available diagnosis for 86.3% of metastatic and 92.0% of primary tumor specimens among 545.",*Pathwork Tissue of Origin Test* matched the known origin of most tumors,86.3% of metastatic and 92.0% of primary tumor specimens,,found,Found by a regulator,tested against a standard,very low,aggregate result,false,Pathwork Tissue of Origin Test,,"Pathwork Diagnostics, Inc.",United States,Pathwork Diagnostics (clinical validation at four processing laboratories),United States,,,Healthcare,,Health,Prediction from data,Patients,3,,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,The available diagnosis for each specimen,Available diagnosis (reference),Agreement for 86.3% of metastatic and 92.0% of primary tumor specimens,measured,persisted,2013-08-26,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/K080896.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, Center for Devices and Radiological Health. 510(k) substantial equivalence determination decision summary for the Pathwork Tissue of Origin Test (K080896). Cleared 2008-07-30."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/K080896.pdf""}]","{""kind"": ""Official documents""}" @@ -349,7 +349,7 @@ AIBD-1125,2025-03-27,source publication,day,mixed,"ESGE's panel weakly recommend AIBD-0143,2025-02-25,event,day,reported,ECMWF's machine-learning forecast model AIFS Single goes operational on 25 February 2025. ECMWF reports lower errors than the physics-based IFS for many measures. Two independent studies in China partly confirm the gain,ECMWF's AIFS Single forecast model went into operation on 25 February 2025. ECMWF reports 5–15% lower errors than IFS. Two teams in China with no tie to ECMWF partly confirm the gain.,ECMWF's AIFS Single goes operational with 5–15% lower errors,5–15% lower errors than IFS,,studied,Independently studied,tested against a standard,low,aggregate result,false,ECMWF Artificial Intelligence Forecasting System (AIFS),AIFS Single v1 (operational from 25 February 2025),ECMWF,United Kingdom,ECMWF,United Kingdom,"Reading, United Kingdom",United Kingdom,Weather forecasting,Public-sector meteorological service,Environment,Prediction from data,,13,High income,environmental_impact,Weather and environmental forecasting,2,3,4,,5,5,,personal; organizational; societal,9,2,ECMWF's physics-based Integrated Forecasting System (IFS),,errors 5 to 15% lower in the medium range (out to 15 days),measured,scaled,2025-07-01,shown,https://journals.ametsoc.org/view/journals/wefo/40/9/WAF-D-24-0227.1.xml,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Deutscher Wetterdienst. Application and verification of ECMWF products 2024: Germany. ECMWF Green Book 2024, section 3.6 b, p. 4. Germany's national weather service verified the AI forecasts against observations. It neither built nor runs them. Germany, which DWD serves, is a member state that funds ECMWF, the developer."", ""url"": ""https://www.ecmwf.int/sites/default/files/elibrary/092024/81600-application-and-verification-of-ecmwf-products-2024-germany.pdf""}]","{""kind"": ""Official documents""}" AIBD-2271,2024-01-30,event,day,reported,Artificial Intelligence-Based Automated Echocardiographic Analysis and the Workflow of Sonographers: A Randomized Crossover Trial (AI-Echo RCT),"In a crossover trial in Tokyo, four sonographers with Us2.ai's automated measurements examined 16.7 patients a day against 14.1, and each examination took 13.0 minutes against 14.3.",*Us2.ai* lets sonographers examine more patients with less fatigue,13.0 minutes,,studied,Independently studied,randomized trial,moderate,aggregate result,false,Us2.ai platform (automated echocardiographic measurements),,Us2.ai (Singapore),Singapore,"Echocardiography laboratory of Juntendo University, Tokyo",Japan,"Tokyo, Japan",Japan,Echocardiography (measurement and reporting),,Health,Images and video,Health workers,3,High income,operational_improvement,Enhanced productivity,2,2,3,,3,5,,organizational; personal,3,0,The manual workflow on days without the platform,14.3 minutes per examination,13.0 minutes per examination,measured,unknown,,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC12909003/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI)-assisted echocardiography analysis and reporting to support the diagnosis and monitoring of heart failure: early-use assessment. HealthTech guidance HTG779, section 1: Recommendations. 19 May 2026."", ""url"": ""https://www.nice.org.uk/guidance/htg779/chapter/1-Recommendations""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI)-assisted echocardiography analysis and reporting to support the diagnosis and monitoring of heart failure: early-use assessment. HealthTech guidance HTG779, section 1: Recommendations. 19 May 2026."", ""url"": ""https://www.nice.org.uk/guidance/htg779/chapter/1-Recommendations""}]","{""kind"": ""Official documents""}" AIBD-0737,2024-01-23,source publication,day,mixed,"NICE allows e-Stroke, RapidAI and Viz in NHS stroke care while evidence is gathered, though it cannot tell how much they speed treatment","In studies that NICE reviewed, stroke treatment came faster after centers adopted e-Stroke, RapidAI or Viz. Treatment rates rose. NICE could not tell how much was due to the software.","Faster stroke treatment followed *e-Stroke*, *RapidAI* and *Viz*",,,studied,Independently studied,before and after,very low,aggregate result,false,"e-Stroke, RapidAI and Viz (AI-derived software for CT brain scans in suspected stroke)",,"Brainomix (e-Stroke), iSchemaView (RapidAI) and Viz.ai (Viz)","Multiple: United Kingdom, United States","NHS stroke centers in England, with AI decision support live at 99 of 107 stroke units",United Kingdom,"England, United Kingdom",United Kingdom,Healthcare (acute stroke care),,Health,Images and video,Patients,3,High income,physical_health,Medical access,2,,1,,2,1,,personal,4,0,"Time to treatment before centers adopted the software, in seven observational studies",,Shorter time to thrombectomy or thrombolysis in most of the studies,measured,persisted,2025-12-02,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN170073.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI)-derived software to help clinical decision making in stroke. HealthTech guidance HTG708 (formerly DG57), section 1: Recommendations. 23 January 2024, last updated 2 May 2024."", ""url"": ""https://www.nice.org.uk/guidance/htg708/chapter/1-Recommendations""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. HTG708, section 3: Committee discussion. 23 January 2024, last updated 2 May 2024."", ""url"": ""https://www.nice.org.uk/guidance/htg708/chapter/3-Committee-discussion""}, {""label"": ""Assessment report"", ""value"": ""Agència de Qualitat i Avaluació Sanitàries de Catalunya (AQuAS). Evaluación de la utilidad diagnóstica, la seguridad, la eficacia y efectividad clínica, la eficiencia y los aspectos técnicos y organizativos de los softwares de inteligencia artificial (IA) para el análisis de imágenes de perfusión cerebral en el ictus isquémico agudo de inicio indeterminado o del despertar. Informes de Evaluación de Tecnologías Sanitarias, Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del Sistema Nacional de Salud. Ministerio de Sanidad, 2026. The quotations are in Spanish."", ""url"": ""https://scientiasalut.gencat.cat/bitstream/11351/14344/4/evaluacion-softwares-inteligencia-artificial-ia-analisis-imagenes-perfusion-cerebral-ictus-sistemico-agudo-2026.pdf""}, {""label"": ""Assessment report"", ""value"": ""Agència de Qualitat i Avaluació Sanitàries de Catalunya (AQuAS). Evaluación de la utilidad diagnóstica, la seguridad, la eficacia y efectividad clínica, la eficiencia y los aspectos técnicos y organizativos de los softwares de inteligencia artificial (IA) para el análisis de imágenes de perfusión cerebral en el ictus isquémico agudo de inicio indeterminado o del despertar. Informes de Evaluación de Tecnologías Sanitarias, Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del Sistema Nacional de Salud. Ministerio de Sanidad, 2026. The quotations are in Spanish."", ""url"": ""https://scientiasalut.gencat.cat/bitstream/11351/14344/4/evaluacion-softwares-inteligencia-artificial-ia-analisis-imagenes-perfusion-cerebral-ictus-sistemico-agudo-2026.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0184,2022-05-01,event,month,reported,A machine learning model that flags risky Universal Credit advance requests saved the UK benefits department an estimated £4.4 million from 2022–23 to 2024–25,A machine-learning model at the Department for Work and Pensions flags risky Universal Credit advance requests for caseworkers. The department estimates that it saved £4.4 million in three years.,"UK model flags risky benefit advances, saving an estimated £4.4 million",DWP estimates that the model saved £4.4 million from 2022–23 to 2024–25,,studied,Independently studied,,very low,aggregate result,false,Universal Credit advances machine learning model,,Department for Work & Pensions (DWP),United Kingdom,Department for Work & Pensions (DWP),United Kingdom,Great Britain,United Kingdom,Benefit fraud and error,Universal Credit advances,Finance,Prediction from data,Public servants,16,High income,privacy_protection,Fraud detection,2,1,4,2,4,5,0.405,organizational; societal,6,2,a randomized control group sample of advance requests,fraud-risk identification rate of the random control sample (not quantified),around three times more effective at identifying fraud risk,measured,persisted,2026-07-09,shown,https://www.gov.uk/algorithmic-transparency-records/dwp-universal-credit-advances-model,£4.4 million,"advances that would otherwise have been overpaid, saved from 2022-23 to 2024-25",True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Inquiry report"", ""value"": ""House of Commons Committee of Public Accounts (UK) (2023). The Department for Work & Pensions Annual Report and Accounts 2022-23 (Fourth Report of Session 2023-24, HC 290), Paragraph 32 (section 'Using machine learning to flag potentially fraudulent benefit claims') and conclusion 6."", ""url"": ""https://web.archive.org/web/20240411110206id_/https://publications.parliament.uk/pa/cm5804/cmselect/cmpubacc/290/report.html""}]","{""kind"": ""Official documents""}" +AIBD-0184,2022-05-01,event,month,reported,A machine learning model that flags risky Universal Credit advance requests saved the UK benefits department an estimated £4.4 million from 2022–23 to 2024–25,A machine-learning model at the Department for Work and Pensions flags risky Universal Credit advance requests for caseworkers. The department estimates that it saved £4.4 million in three years.,"UK model flags risky benefit advances, saving an estimated £4.4 million",DWP estimates that the model saved £4.4 million from 2022–23 to 2024–25,,studied,Independently studied,no comparison,very low,aggregate result,false,Universal Credit advances machine learning model,,Department for Work & Pensions (DWP),United Kingdom,Department for Work & Pensions (DWP),United Kingdom,Great Britain,United Kingdom,Benefit fraud and error,Universal Credit advances,Finance,Prediction from data,Public servants,16,High income,privacy_protection,Fraud detection,2,1,4,2,4,5,0.405,organizational; societal,6,2,a randomized control group sample of advance requests,fraud-risk identification rate of the random control sample (not quantified),around three times more effective at identifying fraud risk,measured,persisted,2026-07-09,shown,https://www.gov.uk/algorithmic-transparency-records/dwp-universal-credit-advances-model,£4.4 million,"advances that would otherwise have been overpaid, saved from 2022-23 to 2024-25",True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Inquiry report"", ""value"": ""House of Commons Committee of Public Accounts (UK) (2023). The Department for Work & Pensions Annual Report and Accounts 2022-23 (Fourth Report of Session 2023-24, HC 290), Paragraph 32 (section 'Using machine learning to flag potentially fraudulent benefit claims') and conclusion 6."", ""url"": ""https://web.archive.org/web/20240411110206id_/https://publications.parliament.uk/pa/cm5804/cmselect/cmpubacc/290/report.html""}]","{""kind"": ""Official documents""}" AIBD-2272,2022-02-22,source publication,day,reported,Effect of Artificial Intelligence Tutoring vs Expert Instruction on Learning Simulated Surgical Skills Among Medical Students: A Randomized Clinical Trial,"In a Montreal trial of 70 medical students learning a simulated brain tumor operation, an AI tutor raised practice performance scores by 0.66 points over remote expert instruction.",*Virtual Operative Assistant* raises scores in simulated surgery,0.66 points,,studied,Independently studied,randomized trial,low,aggregate result,false,"Virtual Operative Assistant (VOA), an AI tutor for virtual reality surgical simulation",,"Neurosurgical Simulation and Artificial Intelligence Learning Centre, McGill University",Canada,"Neurosurgical Simulation and Artificial Intelligence Learning Centre, McGill University (trial with students of four institutions)",Canada,"Montreal, Quebec, Canada",Canada,Surgical education in simulation,,Education,Prediction from data,Students,3; 4,High income,knowledge_discovery,Increased/personalized knowledge,2,2,1,,3,1,,personal,2,0,Remote expert instruction,,0.66 points higher practice score,measured,challenged,2023-09-05,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC8864513/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""A follow-up analysis of the trial by the same group, published in September 2023, found that the tutor's feedback also slowed the students' dominant hand and lowered their rate of tumor removal, which moved these measures away from the experts' benchmarks. Later trials of the group paired AI tutoring with human instructors."", ""url"": ""https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=37725373&rettype=abstract&retmode=text""}]","{""kind"": ""Challenges""}" AIBD-0636,2019-03-01,event,month,reported,HYPE randomized trial at Amsterdam UMC: The Hypotension Prediction Index with a treatment protocol cuts intraoperative hypotension,"In the HYPE trial at Amsterdam UMC, the Hypotension Prediction Index cut median time in hypotension during surgery from 32.7 to 8.0 minutes per patient.",An early warning helps Amsterdam anesthetists keep blood pressure up,median time in hypotension fell from 32.7 to 8.0 minutes per patient,,studied,Independently studied,randomized trial,low,aggregate result,false,Hypotension Prediction Index (HPI),,Edwards Lifesciences,United States,"Amsterdam UMC, location AMC",Netherlands,"Amsterdam UMC, location AMC, Amsterdam, the Netherlands",Netherlands,Healthcare,,Health,Prediction from data,Patients,3,High income,physical_health,,2,2,1,,5,1,,personal,7,3,Standard care with the same blood pressure goal,32.7 minutes median time in hypotension per patient,8.0 minutes median time in hypotension per patient,measured,challenged,2020-12-01,shown,https://www.nice.org.uk/advice/mib266/chapter/Clinical-and-technical-evidence,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration. De Novo Classification Request for Acumen Hypotension Prediction Index (HPI) Feature Software, DEN160044, decision summary, 16 March 2018, pp. 3 and 7. The FDA authorized the software on retrospective accuracy, which does not measure the clinical benefit in this record."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN160044.pdf""}, {""label"": ""Follow-up"", ""value"": ""In a pilot randomized trial in moderate- to high-risk noncardiac surgery, guidance from the index produced no reduction in hypotension."", ""url"": ""https://pubmed.ncbi.nlm.nih.gov/32960954/""}]","{""kind"": ""Challenges""}" AIBD-0295,2018-05-01,event,month,reported,"Dexcom G6's “Urgent Low Soon” predictive alert may cut time below 54 mg/dL, by less in controlled trials","Dexcom staff link the G6's Urgent Low Soon alert to less hypoglycemia in 1,424 users. A second cohort had 15,000. In a Dexcom-funded trial in 24 adults, the reduction appeared only below 2.8 mmol/L.",Dexcom G6's Urgent Low Soon alert may cut time below 54 mg/dL,time below 54 mg/dL fell from 1.0% to 0.6%,,studied,Independently studied,before and after,very low,aggregate result,false,"Dexcom G6 real-time continuous glucose monitoring system, 'Urgent Low Soon' predictive low glucose alert (also present on the G7 in one independent study)",Dexcom G6 (vendor analyses; PACE trial); Dexcom G7 (Preechasuk et al.),"Dexcom, Inc.",United States,,,"United States (vendor analyses); London, United Kingdom (Imperial College trials)",United States | United Kingdom,Healthcare (diabetes technology),Diabetes technology,Health,Prediction from data,Patients,3,High income,physical_health,,2,4,4,3,5,5,0.650,personal; organizational,11,5,"the same users on the preceding Dexcom G5 system, which had no predictive alert (before-and-after)","1.0 (1.5) percent of time <54 mg/dL on G5 (70 mg/dL low-alert setting, Table 1)","0.6 (1.0) percent of time <54 mg/dL on G6 (70 mg/dL low-alert setting, Table 1)",measured,persisted,2022-09-08,not shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC6477579/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration (2018). Evaluation of Automatic Class III Designation for Dexcom G6 Continuous Glucose Monitoring System, Decision Summary, DEN170088 (decision date 27 March 2018)."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN170088.pdf""}]","{""kind"": ""Official documents""}" @@ -357,13 +357,13 @@ AIBD-0058,2016-07-01,event,month,reported,Stanford's Superpower Glass gives chil AIBD-0947,2026-10-06,source publication,day,mixed,ADAC finds in a two-hour test drive in Denmark that Tesla's FSD Supervised drove well over long stretches but still made driving errors,"In a two-hour drive in Denmark, ADAC found that Tesla's FSD Supervised drove well over long stretches but made errors that would have failed it twice in a German driving test.",ADAC finds that *FSD Supervised* drives well in Denmark but still errs,,,studied,Independently studied,no comparison,very low,benefit created,false,"Tesla FSD Supervised (renamed Tesla Assisted Driving), on a Tesla Model 3",,Tesla,United States,"Tesla, under a special permit in Denmark",United States,"Copenhagen, Denmark",Denmark,Road vehicle assisted driving,,Safety and rescue,Machines that act,Drivers and travelers,11,High income,physical_health,,1,1,1,,2,1,,personal,1,0,,,,,challenged,2026-10-12,shown,https://web.archive.org/web/20201022181519/https://www.tesla.com/autopilotAI,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""ADAC reported on 12 October 2026 that Tesla responded to the criticism of FSD Supervised by renaming it Tesla Assisted Driving, which makes clearer that it is an assistance system and not an autonomous one."", ""url"": ""https://www.adac.de/rund-ums-fahrzeug/autokatalog/marken-modelle/tesla/tesla-fsd-supervised-testfahrt/""}]","{""kind"": ""Challenges""}" AIBD-1129,2026-07-28,source publication,day,reported,"Which? finds that Apple's Live Translation on AirPods Pro 3 was generally accurate in Italian, Spanish, German and French conversations","In tests published on 28 July 2026, Which? found that Apple's Live Translation on AirPods Pro 3 was generally accurate in Italian, Spanish, German and French conversations.",*Live Translation* on AirPods is generally accurate in tests by Which?,,,studied,Independently studied,no comparison,very low,aggregate result,false,"Live Translation, as tested on AirPods Pro 3 with an iPhone",,Apple,United States,Apple (feature of iOS 26 for AirPods and iPhone),United States,United Kingdom (Which? tests),United Kingdom,Consumer technology,,Everyday technology,Speech and sound; Text and language,Consumers,,High income,accessibility,,1,,1,,3,,,personal,3,0,,,,,persisted,2026-09-17,shown,https://support.apple.com/guide/iphone/translate-messages-calls-and-conversations-iph22b72984d/26/ios/26,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Testaankoop (expertise Richard Motte, editing Evelien De Bal). Beste vertaal oortjes: onze test van translator earbuds en hun werkelijke prestaties. 8 March 2026. Section on ordinary earbuds with built-in translation, on Apple Live Translate on AirPods Pro 3, tested under iOS 26.1. In the quoted passage, Testaankoop lists instability, an unnatural voice and patchy availability as drawbacks. It concludes that the feature may be stronger in the future but is still too unstable for daily use. Cited as a contrary finding from an earlier test."", ""url"": ""https://www.test-aankoop.be/hightech/koptelefoons-en-oortjes/nieuws/earbuds-vertaal-oortjes""}]","{""kind"": ""Official documents""}" AIBD-0300,2026-05-26,source publication,day,reported,A 79-year-old starts CPAP without polysomnography once ECGs on his Apple Watch Series 11 show atrial fibrillation on mornings after nights of “elevated” breathing disturbance,Apple Watch sleep apnea notifications brought a 79-year-old man in Fukuoka to his cardiologists. His watch ECGs showed atrial fibrillation after nights rated “elevated”. He began CPAP therapy.,Apple Watch breathing alerts send a 79-year-old to CPAP therapy,,,studied,Independently studied,no comparison,very low,benefit created,false,Apple Watch Sleep Apnea Notification Feature (breathing-disturbance notifications) and the ECG app,Apple Watch Series 11 with iPhone 15 Pro,Apple,United States,,,"Fukuoka, Japan",Japan,Healthcare (cardiology; sleep medicine),Consumer wearables,Health,Prediction from data,Patients; Older people,3,High income,physical_health,Medical diagnosis,1,1,,,3,,,personal,3,0,"ambulatory monitoring and standard 12-lead ECG, which had documented no paroxysms since his retirement",AF paroxysms were not documented by ambulatory monitoring or standard ECG,"AF paroxysms were detected repeatedly in the morning after notification of nocturnal “elevated” breathing disturbance, but not in the morning after nights when breathing disturbance was “not elevated”",measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf24/K240929.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Apple Inc. (2024). Sleep Apnea Notification Feature: 510(k) Summary, K240929, device description and performance testing, pp. 2 and 5–7, filed with the FDA. File includes FDA's substantial-equivalence letter of 13 September 2024."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf24/K240929.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0176,2015-01-01,event,year,reported,Some subprime borrowers get lower-priced credit at LendingClub,LendingClub's rating grades relied increasingly on alternative data. They moved some borrowers who are subprime by traditional criteria into better grades with lower-priced credit.,Some subprime borrowers get lower-priced credit at LendingClub,about 8 percent of A-rated and 28 percent of B-rated borrowers had FICO scores below 680,,studied,Independently studied,,very low,aggregate result,false,LendingClub rating-grade model (grades A to G),,LendingClub,United States,LendingClub,United States,United States,United States,Consumer credit,Marketplace consumer lending,Finance,Prediction from data,Consumers,8; 10,High income,economic_access,Fair pricing,1,,4,4,3,3,,personal; societal,3,1,classification of borrowers by traditional credit criteria,classified as subprime by traditional criteria,"slotted into “better” loan grades, which allowed them to get lower-priced credit",measured,litigated,2021-07-14,shown,https://web.archive.org/web/2026id_/https://www.sec.gov/Archives/edgar/data/1409970/000140997020000212/a201910-k.htm,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Follow-up"", ""value"": ""The FTC sued LendingClub in April 2018 over up-front origination fees hidden from loan applicants, and LendingClub settled for $18 million in July 2021; this bears on the full price borrowers paid, not on the rating grades themselves."", ""url"": ""https://web.archive.org/web/2026id_/https://www.ftc.gov/news-events/news/press-releases/2021/07/lendingclub-agrees-pay-18-million-settle-ftc-charges""}]","{""kind"": ""Legal proceedings""}" +AIBD-0176,2015-01-01,event,year,reported,Some subprime borrowers get lower-priced credit at LendingClub,LendingClub's rating grades relied increasingly on alternative data. They moved some borrowers who are subprime by traditional criteria into better grades with lower-priced credit.,Some subprime borrowers get lower-priced credit at LendingClub,about 8 percent of A-rated and 28 percent of B-rated borrowers had FICO scores below 680,,studied,Independently studied,tested against a standard,low,aggregate result,false,LendingClub rating-grade model (grades A to G),,LendingClub,United States,LendingClub,United States,United States,United States,Consumer credit,Marketplace consumer lending,Finance,Prediction from data,Consumers,8; 10,High income,economic_access,Fair pricing,1,,4,4,3,3,,personal; societal,3,1,classification of borrowers by traditional credit criteria,classified as subprime by traditional criteria,"slotted into “better” loan grades, which allowed them to get lower-priced credit",measured,litigated,2021-07-14,shown,https://web.archive.org/web/2026id_/https://www.sec.gov/Archives/edgar/data/1409970/000140997020000212/a201910-k.htm,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Follow-up"", ""value"": ""The FTC sued LendingClub in April 2018 over up-front origination fees hidden from loan applicants, and LendingClub settled for $18 million in July 2021; this bears on the full price borrowers paid, not on the rating grades themselves."", ""url"": ""https://web.archive.org/web/2026id_/https://www.ftc.gov/news-events/news/press-releases/2021/07/lendingclub-agrees-pay-18-million-settle-ftc-charges""}]","{""kind"": ""Legal proceedings""}" AIBD-0390,2025-12-05,source publication,day,tested,"Dentbird software designs dental crowns faster in university tests, but Ohio State researchers conclude that its designs need human review","In university tests, Imagoworks' Dentbird software let technicians design crowns about four times faster. Ohio State researchers concluded that its designs needed human review.",Dentbird software designs crowns faster but still needs human review,,,studied,Independently studied,controlled comparison,very low,aggregate result,false,Dentbird Solutions,,Imagoworks Inc.,South Korea,,,,,Healthcare,,Health,Images and video,Other workers,3,,operational_improvement,Enhanced productivity,,2,1,3,,1,,organizational; personal,7,0,Crowns designed by the same five technicians in conventional CAD software (3Shape Dental System),397.2 ± 80.4 min for the premolar crown with conventional CAD,99.4 ± 17.1 min for the premolar crown with Dentbird,measured,challenged,2025-12-05,shown,https://www.dental-tribune.com/news/imagoworks-propels-digital-dentistry-into-the-ai-era-at-ids-2023/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""Ohio State researchers found that Dentbird saved time on three-unit bridges but underperformed on esthetic and functional criteria, and they concluded that human oversight is still needed. Their April 2025 study of single crowns reached a similar result."", ""url"": ""https://pubmed.ncbi.nlm.nih.gov/41353038/""}]","{""kind"": ""Challenges""}" AIBD-1162,2025-04-30,source publication,day,tested,Singapore's ACE finds that EchoGo Heart Failure reclassified most cases that clinical scores left indeterminate,Singapore's ACE found that EchoGo Heart Failure appeared safe on retrospective data. It appeared accurate. It reclassified 73.5% of cases left indeterminate by the HFA-PEFF score.,*EchoGo Heart Failure* reclassified most cases that scores left open,reclassified 73.5% of cases left indeterminate by the HFA-PEFF score,,studied,Independently studied,tested against a standard,very low,aggregate result,false,EchoGo Heart Failure,,Ultromics Ltd.,United Kingdom,Hospitals in the retrospective US and UK studies that ACE reviewed (not yet in use in Singapore),United States and United Kingdom,"United States and United Kingdom (studies), assessed for Singapore",United States | United Kingdom | Singapore,Healthcare,,Health,Images and video,Health workers,3,High income,physical_health,Medical diagnosis,,,,,,,,organizational,2,0,The HFA-PEFF and H2FPEF clinical scores in the main training and validation study,54.6% and 60.4% non-diagnostic outcomes,7.3% non-diagnostic outputs,measured,unknown,,shown,https://isomer-user-content.by.gov.sg/68/55e34a4a-9a64-4e9d-9fe9-dea47842f372/EchoGo%20Heart%20Failure%20to%20Aid%20in%20the%20Diagnosis%20of%20Heart%20Failure%20with%20Preserved%20Ejection%20Fraction.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI)-assisted echocardiography analysis and reporting to support the diagnosis and monitoring of heart failure: early-use assessment. HealthTech guidance HTG779, section 1: Recommendations. 19 May 2026."", ""url"": ""https://www.nice.org.uk/guidance/htg779/chapter/1-Recommendations""}, {""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence. Artificial intelligence (AI)-assisted echocardiography analysis and reporting to support the diagnosis and monitoring of heart failure: early-use assessment. HealthTech guidance HTG779, section 1: Recommendations. 19 May 2026."", ""url"": ""https://www.nice.org.uk/guidance/htg779/chapter/1-Recommendations""}]","{""kind"": ""Official documents""}" AIBD-0764,2025-01-02,source publication,day,tested,University of Geneva researchers find that GraphCast and Pangu-Weather forecast the 2021 North American winter storm better than ECMWF's HRES,"In three case studies, GraphCast and Pangu-Weather forecast the February 2021 North American winter storm better than ECMWF's HRES. They did less well on two heatwaves.",*GraphCast* and *Pangu-Weather* beat ECMWF on the 2021 winter storm,,,studied,Independently studied,tested against a standard,very low,aggregate result,false,"GraphCast, Pangu-Weather and FourCastNet",,"Google DeepMind (GraphCast), Huawei (Pangu-Weather) and NVIDIA (FourCastNet)","Multiple: United Kingdom, China, United States",,,"Texas and the southern United States (2021 winter storm), the Pacific Northwest (2021 heatwave) and South Asia (2023 humid heatwave)",United States,Weather forecasting / extreme events,,Environment,Prediction from data,Researchers,13,High income,environmental_impact,Natural disaster prediction,,,,,,,,organizational,1,0,ECMWF HRES forecasts for the same events and lead times,"HRES errors in wind chill near College Station, Texas","Much lower errors for Pangu-Weather and GraphCast, especially from 9 to 17 February",measured,challenged,2025-08-21,shown,https://journals.ametsoc.org/view/journals/aies/4/1/AIES-D-24-0033.1.xml,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""In August 2025 two of the study's authors and colleagues reported that ECMWF's HRES consistently beat GraphCast, Pangu-Weather and other AI models on unprecedented heat, cold and wind extremes across nearly all lead times."", ""url"": ""https://arxiv.org/abs/2508.15724""}]","{""kind"": ""Challenges""}" AIBD-0763,2024-11-07,source publication,day,tested,Uppsala University researchers find that Pangu-Weather and GraphCast mostly beat ECMWF's deterministic model on global error in 2020 forecasts,"In 2020 forecasts, Pangu-Weather and GraphCast mostly beat ECMWF's IFS HRES on global error 1 to 10 days ahead. They matched it on temperature extremes in most regions. The same held for wind.",*Pangu-Weather* and *GraphCast* rival ECMWF on weather extremes,,,studied,Independently studied,tested against a standard,very low,aggregate result,false,Pangu-Weather and GraphCast (operational versions),,Huawei (Pangu-Weather) and Google DeepMind (GraphCast),"Multiple: China, United Kingdom",,,"Global, with the regions of ECMWF's operational scorecards",,Weather forecasting / temperature and wind extremes,,Energy and infrastructure,Prediction from data,Researchers,13,,environmental_impact,Weather and environmental forecasting,,,,,,,,organizational,2,0,ECMWF IFS HRES forecasts from the same initial conditions,IFS HRES errors in the tropics and the North Pacific,Lower GraphCast errors there for all variables and lead times,measured,challenged,2025-08-21,shown,https://gmd.copernicus.org/articles/17/7915/2024/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""A Geneva-led preprint of August 2025 found that ECMWF's HRES consistently beat GraphCast and Pangu-Weather on unprecedented heat, cold and wind extremes at nearly all lead times."", ""url"": ""https://arxiv.org/pdf/2508.15724""}]","{""kind"": ""Challenges""}" -AIBD-0261,2023-11-29,event,day,tested,"GNoME predicts 2.2 million crystal structures, of which few may be useful new compounds","DeepMind's GNoME graph networks proposed 2.2 million crystal structures (380,000 stable). Independent researchers found duplicates. They found scant evidence of useful new compounds.","GNoME predicts 2.2 million crystals, among which critics find duplicates","2.2 million crystal structures (380,000 stable)",,studied,Independently studied,,very low,aggregate result,false,GNoME (Graph Networks for Materials Exploration),,DeepMind,United Kingdom,DeepMind (dataset released openly; stable structures contributed to the Materials Project),United Kingdom,"Mountain View, California, USA",United States,Materials science,Materials researchers,Science,Prediction from data,Researchers,9,High income,knowledge_discovery,,,1,4,,,1,,personal; organizational; societal,11,3,stable crystals previously identified by computational studies before GNoME,"48,000 computationally stable materials","a total of 421,000 stable crystals",measured,challenged,2026-10-09,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC10700131/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""Independent chemists and mathematicians challenged the claims from 2024 to 2026. The Nature paper carries no correction or editorial note, and DeepMind told 404 Media that it stands by its claims, so the benefit shows a challenge and no reversal."", ""url"": ""https://arxiv.org/abs/2610.13885""}]","{""kind"": ""Challenges""}" -AIBD-0071,2021-04-21,source publication,day,tested,"An AI analysis of handwriting divides the Great Isaiah Scroll (1QIsaa) into two groups of columns, which its authors attribute to two scribes, but researchers in Norway question the method in 2025","An AI method from Groningen split the 54 columns of the Great Isaiah Scroll into two groups of handwriting, which its authors attribute to two scribes. Researchers in Norway questioned it in 2025.",AI detects two groups of handwriting in the Great Isaiah Scroll,"two statistically distinct groups, with a transition in columns 27–29",,studied,Independently studied,,very low,benefit created,false,Groningen AI-based writer-identification pipeline (neural-network binarization with textural and allographic feature analysis),BiNet,"University of Groningen (Dhali, Schomaker, Popović)",Netherlands,University of Groningen,Netherlands,"Groningen, Netherlands",Netherlands,History and archives,Palaeography of the Dead Sea Scrolls,Culture and history,Images and video,Researchers,11,High income,knowledge_discovery,,,,5,,,1,,organizational; societal,5,0,the prevailing scholarly view that a single scribe copied the whole scroll,"the entire manuscript was copied by one scribe, with minor interventions by other, contemporaneous and also much later, scribes","two main scribes, each showing different writing patterns, were responsible for the Great Isaiah Scroll",measured,challenged,2025-06-24,shown,https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0249769,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""Tabita Tobing and Patrick Bours of the Norwegian University of Science and Technology retraced a column-based method on the same images in June 2025. They argue that the 2021 analysis was framed to validate a two-scribe hypothesis. They found that the two-scribe split separates letters better than other groupings, with only moderate scores, and they judge that hinge features alone cannot attribute scribes reliably."", ""url"": ""https://library.imaging.org/admin/apis/public/api/ist/website/downloadArticle/archiving/22/1/14""}]","{""kind"": ""Challenges""}" +AIBD-0261,2023-11-29,event,day,tested,"GNoME predicts 2.2 million crystal structures, of which few may be useful new compounds","DeepMind's GNoME graph networks proposed 2.2 million crystal structures (380,000 stable). Independent researchers found duplicates. They found scant evidence of useful new compounds.","GNoME predicts 2.2 million crystals, among which critics find duplicates","2.2 million crystal structures (380,000 stable)",,studied,Independently studied,no comparison,very low,aggregate result,false,GNoME (Graph Networks for Materials Exploration),,DeepMind,United Kingdom,DeepMind (dataset released openly; stable structures contributed to the Materials Project),United Kingdom,"Mountain View, California, USA",United States,Materials science,Materials researchers,Science,Prediction from data,Researchers,9,High income,knowledge_discovery,,,1,4,,,1,,personal; organizational; societal,11,3,stable crystals previously identified by computational studies before GNoME,"48,000 computationally stable materials","a total of 421,000 stable crystals",measured,challenged,2026-10-09,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC10700131/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""Independent chemists and mathematicians challenged the claims from 2024 to 2026. The Nature paper carries no correction or editorial note, and DeepMind told 404 Media that it stands by its claims, so the benefit shows a challenge and no reversal."", ""url"": ""https://arxiv.org/abs/2610.13885""}]","{""kind"": ""Challenges""}" +AIBD-0071,2021-04-21,source publication,day,tested,"An AI analysis of handwriting divides the Great Isaiah Scroll (1QIsaa) into two groups of columns, which its authors attribute to two scribes, but researchers in Norway question the method in 2025","An AI method from Groningen split the 54 columns of the Great Isaiah Scroll into two groups of handwriting, which its authors attribute to two scribes. Researchers in Norway questioned it in 2025.",AI detects two groups of handwriting in the Great Isaiah Scroll,"two statistically distinct groups, with a transition in columns 27–29",,studied,Independently studied,no comparison,very low,benefit created,false,Groningen AI-based writer-identification pipeline (neural-network binarization with textural and allographic feature analysis),BiNet,"University of Groningen (Dhali, Schomaker, Popović)",Netherlands,University of Groningen,Netherlands,"Groningen, Netherlands",Netherlands,History and archives,Palaeography of the Dead Sea Scrolls,Culture and history,Images and video,Researchers,11,High income,knowledge_discovery,,,,5,,,1,,organizational; societal,5,0,the prevailing scholarly view that a single scribe copied the whole scroll,"the entire manuscript was copied by one scribe, with minor interventions by other, contemporaneous and also much later, scribes","two main scribes, each showing different writing patterns, were responsible for the Great Isaiah Scroll",measured,challenged,2025-06-24,shown,https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0249769,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""Tabita Tobing and Patrick Bours of the Norwegian University of Science and Technology retraced a column-based method on the same images in June 2025. They argue that the 2021 analysis was framed to validate a two-scribe hypothesis. They found that the two-scribe split separates letters better than other groupings, with only moderate scores, and they judge that hinge features alone cannot attribute scribes reliably."", ""url"": ""https://library.imaging.org/admin/apis/public/api/ist/website/downloadArticle/archiving/22/1/14""}]","{""kind"": ""Challenges""}" AIBD-0221,2020-12-02,source publication,day,tested,"The online screener Dytective detects 80.6% of children with dyslexia among 3,644 participants, though two re-analyses in 2026 show a demographic confound","Dytective is a gamified online test for dyslexia with a random-forest classifier. It screened 3,644 children with 79.8% accuracy. Two papers in 2026 showed a demographic confound in its dataset.",Dytective detects 80.6% of dyslexia cases in a study later challenged,79.8% accuracy with 80.6% recall and 79.3% precision,,studied,Independently studied,tested against a standard,very low,aggregate result,false,Dytective online gamified dyslexia screening test with a random-forest classifier,Weka 3.8.3 implementation with 200 trees and unlimited height,"Luz Rello and co-authors (IE University, Northeastern, University of Washington, Carnegie Mellon, University of Barcelona), per the paper's author list",,"Schools, dyslexia centers and non-profits in Argentina, Chile, Colombia, Spain and the USA (study participants); public online use in Spanish-speaking countries",,"Argentina, Chile, Colombia, Spain and the United States (study participants)",Argentina | Chile | Colombia | Spain | United States,Education / learning-disability screening,School-age dyslexia screening,Accessibility,Prediction from data,Children; People with disabilities,4,Upper middle income; High income,incentivization,,,,,,,,,personal; organizational,4,1,a trivial classifier that labels every participant as not having dyslexia,89.2% accuracy with no child with dyslexia detected,79.8% accuracy with 80.6% recall and 79.3% precision for the dyslexia class,measured,challenged,2026-09-23,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC7710040/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""Two independent papers by Sudha and Murali (2 September and 23 September 2026) re-analyzed the public dataset. They report that four demographic columns carry much of the discrimination and that false-positive rates differ twelve-fold between language groups. They do not test the deployed tool on new children, so the screener's results are questioned on this dataset without a finding that the tool fails in use."", ""url"": ""https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2026.1927725/full""}]","{""kind"": ""Challenges""}" AIBD-0066,2020-09-01,event,month,reported,The developers of TREWS report that sepsis patients whose alert was confirmed within three hours had lower mortality at five hospitals,"In a five-hospital study by the TREWS developers, patients whose sepsis alert was confirmed within three hours had 3.3% absolute lower in-hospital mortality, a link that outside researchers question.",Confirmed TREWS sepsis alerts are linked to 3.3% lower mortality,3.3% absolute (18.7% relative) lower in-hospital mortality,,studied,Independently studied,controlled comparison,very low,aggregate result,false,TREWS (Targeted Real-time Early Warning System),,Johns Hopkins University (Suchi Saria's group),United States,Johns Hopkins Medicine (five hospitals),United States,"Baltimore, United States",United States,Healthcare - sepsis / critical care,EHR-integrated early warning system,Health,Prediction from data,Patients,3,High income,physical_health,Medical diagnosis,,3,4,,3,5,,personal; societal,9,0,alerted sepsis patients whose alert was not confirmed by a provider within 3 h,patients whose alert was not confirmed by a provider within 3 h,"reduced in-hospital mortality rate (3.3%, confidence interval (CI) 1.7, 5.1%, adjusted absolute reduction, and 18.7%, CI 9.4, 27.0%, adjusted relative reduction)",measured,scaled,2026-08-04,shown,https://undark.org/2022/10/12/a-new-transparent-ai-tool-may-help-detect-blood-poisoning/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""Bayesian Health, 510(k) summary for the Bayesian Health Sepsis Flagging Device (K250680), with the FDA clearance letter of 30 April 2026. Bayesian Health holds a license agreement with Johns Hopkins and wrote the summary. The FDA found substantial equivalence to an earlier sepsis device and made no finding on mortality."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf25/K250680.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Centers for Medicare & Medicaid Services. Medicare Program; Hospital Inpatient Prospective Payment Systems for Acute Care Hospitals and Fiscal Year 2027 Rates (final rule). Federal Register 91 FR 49570, 4 August 2026."", ""url"": ""https://www.federalregister.gov/documents/2026/08/04/2026-15833/medicare-program-hospital-inpatient-prospective-payment-systems-for-acute-care-hospitals-ipps-and""}, {""label"": ""Regulatory document"", ""value"": ""Centers for Medicare & Medicaid Services. Medicare Program; Hospital Inpatient Prospective Payment Systems for Acute Care Hospitals and Fiscal Year 2027 Rates (final rule). Federal Register 91 FR 49570, 4 August 2026."", ""url"": ""https://www.federalregister.gov/documents/2026/08/04/2026-15833/medicare-program-hospital-inpatient-prospective-payment-systems-for-acute-care-hospitals-ipps-and""}]","{""kind"": ""Official documents""}" AIBD-1104,2020-06-01,event,month,tested,Consumer Reports finds that Tesla's Autopilot on a Model Y did the best lane keeping of 17 driving assistance systems in its 2020 tests,"Consumer Reports tested 17 driving assistance systems in 2020. Tesla's Autopilot, on a Model Y, did the best lane keeping. It scored 9 of 10 for capability and performance, the top score.",*Autopilot* did the best lane keeping of 17 systems in CR's 2020 tests,9 of 10 for capability and performance,,studied,Independently studied,tested against a standard,very low,aggregate result,false,"Autopilot, as fitted to the Tesla Model Y (2020)",,Tesla,United States,Tesla (fitted to the Model Y),United States,United States,United States,Road vehicle crash avoidance / assisted driving,,Safety and rescue,Machines that act,Drivers and travelers,11,High income,physical_health,Hazard reduction,,,,,,,,personal,2,0,The 16 other active driving assistance systems in the same Consumer Reports tests,8 of 10 or less for capability and performance,9 of 10,measured,challenged,2024-04-26,shown,https://web.archive.org/web/20201022181519/https://www.tesla.com/autopilotAI,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""In April 2024 NHTSA closed its Autopilot investigation, citing 13 fatal crashes related to misuse, and said that Tesla's weak driver engagement system was not appropriate for Autopilot's permissive operating capabilities."", ""url"": ""https://techcrunch.com/2024/04/26/tesla-nhtsa-autopilot-investigation-closed-fatal-crashes""}]","{""kind"": ""Challenges""}" @@ -371,7 +371,7 @@ AIBD-0949,2020-01-01,event,year,tested,ADAC reports that Tesla's Autopilot on th AIBD-0328,2026-04-05,event,day,reported,"A CIA tool called “Ghost Murmur”, pairing quantum magnetometry with AI, may have located downed F-15E officer “Dude 44 Bravo” in Iran before his rescue in April 2026, but CBS later reports that unnamed CIA technology pinpointed him","Two anonymous sources credit the CIA's Ghost Murmur, quantum sensing paired with AI, with locating a downed F-15E officer in Iran. CBS reports that CIA technology pinpointed him. It names no tool.",*Ghost Murmur* may have located a downed F-15E officer in Iran,"one rescued airman, with the claimed AI contribution unverified",The rescue of a downed American airman in Iran is established by the White House briefing of 6 April.[1] The claim that an AI-assisted quantum sensor located him comes from two anonymous sources in the New York Post. President Trump later told the Post that the tool was important but did not explain how it worked.[2] Physicists interviewed by the Post and by Scientific American questioned the capability as described.[3] The contribution attributed to Ghost Murmur remains unverified.,reported,Independently reported,no comparison,very low,harm prevented,false,"'Ghost Murmur': long-range quantum (nitrogen-vacancy diamond) magnetometry paired with AI software that isolates heartbeat signatures from background noise, as described by the NY Post's anonymous sources",,"Lockheed Martin Skunk Works, per two NY Post sources; the company declined to comment",United States,"US Central Intelligence Agency, per the NY Post; the CIA and White House named no tool",United States,Mountains of southern Iran,Iran,Defense and intelligence: combat search and rescue,Remote personnel-location sensing,Defense,Prediction from data,Public servants,3,Upper middle income,physical_health,,5,1,5,,3,1,,personal; organizational,6,0,,,,,challenged,2026-04-08,not shown,https://nypost.com/2026/04/07/us-news/ghost-murmur-a-never-used-secret-tool-deployed-to-find-lost-airman-in-iran-in-daring-mission/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""Physicists interviewed by the Post and Scientific American disputed the public technical description. The claim about the rescue has not been formally reversed or withdrawn."", ""url"": ""https://www.scientificamerican.com/article/what-is-the-quantum-ghost-murmur-purportedly-used-in-iran-scientists/""}]","{""kind"": ""Challenges""}" AIBD-2229,2025-04-02,source publication,day,reported,Real-time surveillance system for patient deterioration: a pragmatic cluster-randomized controlled trial,"In a cluster-randomized trial in two U.S. health systems, patients on units that used the CONCERN early warning system had a 35.6% lower risk of death in the hospital than patients in usual care.",*CONCERN* alerts from nurses' charting cut hospital deaths in a trial,35.6% lower risk of death,,reported,Independently reported,randomized trial,high,aggregate result,false,CONCERN early warning system,,Columbia University with Brigham and Women's Hospital,United States,Two health systems in the United States,United States,United States,United States,Inpatient care (early warning of deterioration from nursing documentation),,Health,Prediction from data,Patients; Health workers,3,High income,physical_health; operational_improvement,Medical diagnosis,5,4,5,,3,,,personal,3,0,"Clinical units in usual care, by cluster randomization","265 deaths among 27,869 encounters","181 deaths among 33,024 encounters",measured,unknown,,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC12818231/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Winters BD, Rosen M, Sharma R, Zhang A, Bass EB. Patient Monitoring Systems To Prevent Failure To Rescue: Rapid Review. Making Healthcare Safer IV. Agency for Healthcare Research and Quality, December 2024. The review treats early warning systems as a class."", ""url"": ""https://pubmed.ncbi.nlm.nih.gov/39841855/""}]","{""kind"": ""Official documents""}" AIBD-0579,2022-12-07,event,day,reported,Sepsis alerts from COMPOSER are associated with lower mortality in two emergency departments at UC San Diego Health,"After UC San Diego Health deployed COMPOSER, a deep-learning model that alerts nurses to sepsis, mortality was 9.49%, against a modeled 11.39% without it.",Fewer sepsis patients die after COMPOSER alerts UC San Diego nurses,1.9 percentage points,"COMPOSER is a deep-learning model that flags possible sepsis to nurses. It went live in two emergency departments at UC San Diego Health on 7 December 2022. The evaluation covered 6,217 adults with sepsis seen before and after that date. Mortality after deployment was 9.49%, against a modeled 11.39% without COMPOSER. That is an adjusted reduction of 1.9 percentage points, with a confidence interval of 0.3–3.5 at the 95% level. Raw mortality before deployment was 10.3%. The comparison rests on historical controls, with no randomization and no concurrent untreated group. The estimated benefit reached statistical significance at only one of the two hospitals.[1] In a meta-analysis published on 13 May 2025, Yale authors pooled seven randomized trials of other alert systems for sepsis. They found no statistically significant effect on mortality.[2] The COMPOSER team reported in August 2026 that the alerts stayed in use in both emergency departments through October 2025. A large language model was added as a layer in July 2024. The mortality result is an association, reported by the team that built COMPOSER and the health system that deployed it.",reported,Independently reported,no comparison,very low,aggregate result,false,COMPOSER deep-learning sepsis prediction and nurse alert system,,University of California San Diego research team,United States,UC San Diego Health,United States,"San Diego, California, United States",United States,Healthcare - sepsis,Electronic-health-record early warning system,Health,Prediction from data,Patients; Health workers,3,High income,physical_health,Medical diagnosis,5,3,5,2,4,5,0.675,personal,6,1,"Modeled post-deployment mortality without COMPOSER, using historical data and adjustment for confounders",11.39% modeled counterfactual mortality,9.49% observed post-deployment mortality,assumed,persisted,2026-08-05,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC10805720/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Cherian J, Segal J, Sharma R, Zhang A, Bass EB, Rosen M. Patient Safety Practices Focused on Sepsis Prediction and Recognition: Rapid Review. Making Healthcare Safer IV. Agency for Healthcare Research and Quality, 2024. The review treats the practices as a class. Its search ended in August 2023, before the study of COMPOSER appeared."", ""url"": ""https://pubmed.ncbi.nlm.nih.gov/38709865/""}]","{""kind"": ""Official documents""}" -AIBD-0296,2016-11-01,event,month,reported,"A seizure alert from an Empatica Embrace reaches the mother of a 26-year-old teacher, who calls paramedics after a seizure left her daughter not breathing (Fulton, Missouri)","An Embrace wristband detected a seizure that stopped a 26-year-old teacher's breathing. It alerted her mother, who called paramedics. IEEE Spectrum reports that she survived.","Embrace wristband alerts a mother, who calls paramedics",,"In November 2016 a 26-year-old teacher from Fulton, Missouri, had an epileptic seizure. It left her unresponsive and not breathing.[1] Her Embrace wristband detected the seizure and alerted her mother, who called paramedics. The teacher survived. Her rescue rests on a single IEEE Spectrum report. The U.S. Food and Drug Administration cleared the device on 26 January 2018, after the event.[2] Later evidence comes from hospital units where patients with epilepsy are monitored. Researchers in Montreal published a study of the device in their own unit on 7 January 2026. They found that the Embrace2 detected all 16 convulsive seizures that occurred while it was worn and working. It raised 29 false alarms in 373 days of monitoring. It detected none of 510 focal seizures without convulsions, a type that the device is not cleared to detect.[3]",reported,Independently reported,,very low,harm prevented,false,"Empatica Embrace wrist-worn seizure detection (electrodermal activity and accelerometer, on-board algorithm) with the Alert smartphone app",Embrace (the article does not state the generation); echoes concern Embrace2,Empatica,United States,,,"Fulton, Missouri, United States",United States,Healthcare (epilepsy; seizure monitoring),Prescription wearable seizure monitor used at home,Health,Prediction from data,Patients,3,High income,physical_health,,5,1,5,,5,1,,personal,5,3,"seizures recorded during more than 6,500 hours of monitoring of 135 patients (device test reported by IEEE Spectrum)",,"the device correctly identified nearly every single seizure, with a false alarm rate below that of the seizure frequency for most individuals",measured,persisted,2026-01-07,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC8418082/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration (2018). 510(k) Summary K172935, Embrace, Empatica S.r.l., decision 26 January 2018 (text extracted from the PDF)."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf17/K172935.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0296,2016-11-01,event,month,reported,"A seizure alert from an Empatica Embrace reaches the mother of a 26-year-old teacher, who calls paramedics after a seizure left her daughter not breathing (Fulton, Missouri)","An Embrace wristband detected a seizure that stopped a 26-year-old teacher's breathing. It alerted her mother, who called paramedics. IEEE Spectrum reports that she survived.","Embrace wristband alerts a mother, who calls paramedics",,"In November 2016 a 26-year-old teacher from Fulton, Missouri, had an epileptic seizure. It left her unresponsive and not breathing.[1] Her Embrace wristband detected the seizure and alerted her mother, who called paramedics. The teacher survived. Her rescue rests on a single IEEE Spectrum report. The U.S. Food and Drug Administration cleared the device on 26 January 2018, after the event.[2] Later evidence comes from hospital units where patients with epilepsy are monitored. Researchers in Montreal published a study of the device in their own unit on 7 January 2026. They found that the Embrace2 detected all 16 convulsive seizures that occurred while it was worn and working. It raised 29 false alarms in 373 days of monitoring. It detected none of 510 focal seizures without convulsions, a type that the device is not cleared to detect.[3]",reported,Independently reported,no comparison,very low,harm prevented,false,"Empatica Embrace wrist-worn seizure detection (electrodermal activity and accelerometer, on-board algorithm) with the Alert smartphone app",Embrace (the article does not state the generation); echoes concern Embrace2,Empatica,United States,,,"Fulton, Missouri, United States",United States,Healthcare (epilepsy; seizure monitoring),Prescription wearable seizure monitor used at home,Health,Prediction from data,Patients,3,High income,physical_health,,5,1,5,,5,1,,personal,5,3,"seizures recorded during more than 6,500 hours of monitoring of 135 patients (device test reported by IEEE Spectrum)",,"the device correctly identified nearly every single seizure, with a false alarm rate below that of the seizure frequency for most individuals",measured,persisted,2026-01-07,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC8418082/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration (2018). 510(k) Summary K172935, Embrace, Empatica S.r.l., decision 26 January 2018 (text extracted from the PDF)."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf17/K172935.pdf""}]","{""kind"": ""Official documents""}" AIBD-0112,2022-03-20,event,day,reported,ASENA analysis program flags a car at Yüksekova carrying 52.5 kg of explosives bound for Istanbul,The Turkish police's ASENA analysis program flagged a car at Yüksekova on 20 March 2022. Officers found 52.5 kg of explosives. The interior minister said that they were bound for Istanbul.,ASENA flags a car at Yüksekova carrying 52.5 kg of explosives,"133 packets, 52.5 kg, of factory-made explosive","On 20 March 2022 officers of the Hakkari Provincial Police Directorate stopped a car at a checkpoint in Yüksekova, Turkey. Work on the directorate's ASENA analysis program had flagged the car. A hidden compartment under the spare-wheel well held 133 packets of factory-made explosive, 52.5 kg in all. The explosive was assessed as PETN-RDX.[1] The driver was a police officer posted in Istanbul. On 20 November 2023 the Hakkari 2nd Heavy Penal Court sentenced him to 21 years in prison for supplying weapons to a terrorist organization. It sentenced four co-defendants to 17.5 years each for the same offense.[2] ASENA was designed to flag drug crime. This find lay outside its purpose. Interior Minister Süleyman Soylu said on 21 March 2022 that police had identified a suspicious vehicle thought to be carrying drugs. He said that the police had used analysis programs, “what they call artificial intelligence”.[3] The head of the police Narcotics Crimes Department spoke in June 2022. He listed the explosives among the finds of operations that, by his account, had started with ASENA.[4] No source records the data that ASENA matched for this car. None of the sources says whether an analyst checked the flag before the stop. The driver spoke in court of “those who informed”, but no informant is reported.[5]",reported,Independently reported,no comparison,very low,harm prevented,true,"ASENA (Analiz Sistemleri Narkotik Ağı), Turkish National Police narcotics analysis software",,"Turkish National Police (Emniyet Genel Müdürlüğü), Narcotics Crimes Department; described as written by Turkish engineers, no contractor named",Turkey,"Hakkari Provincial Police Directorate, Turkish National Police",Turkey,"Yüksekova, Turkey",Turkey,Public safety and counter-terrorism policing,"Government, security, and defense",Public sector,Prediction from data,Public servants,16,Upper middle income,physical_health,Crime reduction or prevention,4,,4,,3,1,,societal; organizational,8,0,,,,,litigated,2023-11-20,not shown,https://www.aa.com.tr/tr/gundem/yapay-zeka-asena-uyusturucunun-izini-suruyor/2603769,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Follow-up"", ""value"": ""On 20 November 2023 the Hakkari 2nd Heavy Penal Court sentenced the police officer A.K. to 21 years in prison. It sentenced the four co-defendants to 17 years 6 months each."", ""url"": ""https://kisadalga.net/haber/detay/aracinda-52-kilo-patlayici-bulunan-polise-21-yil-hapis-cezasi_86842""}]","{""kind"": ""Legal proceedings""}" AIBD-0239,2026-06-19,event,day,reported,OpenAI alert relayed through the FBI leads Espírito Santo police to detain a father a day before his stated date to kill his son,"OpenAI's safety systems flagged a ChatGPT user's plan to kill his 8-year-old son. Alerted through the FBI, Brazilian police held him a day before his stated date.",OpenAI flag leads Brazilian police to hold a father a day early,held him a day before his stated date,,reported,Independently reported,no comparison,very low,protective response,false,ChatGPT with OpenAI's safety-monitoring pipeline (automated classifiers plus human review and law-enforcement referral),,OpenAI,United States,OpenAI,United States,"São Gabriel da Palha, Brazil",Brazil,Public safety,Law enforcement and child protection,Safety and rescue,Text and language,Children,16,Upper middle income,physical_health,Crime reduction or prevention,3,2,2,,3,1,,personal; organizational,9,1,the date of 20 June 2026 that the user gave in his messages,Police received the tip on 16 June 2026.,"Police arrested him on 19 June 2026, a day before that date.",assumed,litigated,2026-08-13,shown,https://web.archive.org/web/2026id_/https://openai.com/index/helping-people-when-they-need-it-most/,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Follow-up"", ""value"": ""He was released on 13 August 2026 after a court granted habeas corpus, under an ankle monitor and a no-contact order, with the inquiry continuing."", ""url"": ""https://www.bbc.com/portuguese/articles/c8dn1l7vr6ro""}]","{""kind"": ""Legal proceedings""}" AIBD-0260,2026-05-29,event,day,reported,"A man pleads guilty to stalking after OpenAI refers his threats to the FBI, which passes them to local police","Through the FBI, OpenAI disclosed threats that a man wrote in ChatGPT against his ex-girlfriend. He later pleaded guilty to stalking.",A man pleads guilty after OpenAI refers his threats to the FBI,"eight years of probation, including two years of GPS monitoring",,reported,Independently reported,no comparison,very low,protective response,false,"ChatGPT (OpenAI), with OpenAI's automated detection and human review pipeline for users planning to harm others",,OpenAI,United States,OpenAI,United States,"South Palm Beach, United States",United States,Public safety - stalking and threat investigation,Consumer chatbot trust-and-safety referral to law enforcement,Safety and rescue,Text and language,Women and mothers,5; 16,High income,physical_health,Crime reduction or prevention,3,1,4,,3,1,,personal,4,0,"the situation before the referral, when the threatened woman had received anonymous threats but had not reported the man to police",no report to law enforcement; the threats came from burner numbers,detectives matched the anonymous texts to the man through the ChatGPT records and arrested him six days after the FBI notice,assumed,litigated,2026-08-13,shown,https://web.archive.org/web/2026id_/https://openai.com/index/helping-people-when-they-need-it-most/,,,,[],Querist,The use of the system was taken to court.,"[{""label"": ""Follow-up"", ""value"": ""The man pleaded guilty on 13 August 2026 to written threats to kill, aggravated stalking and misuse of a two-way device, with adjudication withheld and eight years of probation."", ""url"": ""https://bocanewsnow.com/2026/08/27/man-flagged-by-chatgpt-pleads-guilty-in-stalking-case/""}]","{""kind"": ""Legal proceedings""}" @@ -387,7 +387,7 @@ AIBD-0447,2024-01-01,event,year,reported,AI over 5G-A networks to protect wildli AIBD-0018,2023-10-19,source publication,day,reported,Convolutional neural network finds 72 previously undetected pulses from fast radio burst FRB 121102 in Green Bank Telescope data,"A convolutional neural network built at the University of California, Berkeley, found 72 missed pulses from the fast radio burst FRB 121102. A search in 2023 found 12 of them again.",Neural network finds 72 missed pulses from a repeating fast radio burst,found 72 previously undetected pulses from the repeating fast radio burst FRB 121102,,reported,Independently reported,tested against a standard,very low,benefit created,false,"Convolutional neural network for FRB pulse detection (Zhang et al., Breakthrough Listen / Berkeley SETI Research Center)",,"University of California, Berkeley (Berkeley SETI Research Center, Breakthrough Listen)",United States,"Breakthrough Listen, on Green Bank Telescope data",United States,"Berkeley, United States",United States,Astrophysics / radio transients,Academic research,Science,Prediction from data,Researchers,,High income,knowledge_discovery,,2,,5,,4,1,,organizational; societal,5,1,the conventional search algorithm previously run on the same observation,21 bursts have been previously detected,"72 new pulses, totaling 93 pulses in five hours",measured,challenged,2023-10-19,shown,https://arxiv.org/abs/1809.03043,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""A new search in 2023 of the first 30 minutes of the same data found 30 of the 46 candidates of the CNN for that span again. Its burst table shows that 18 of the 30 were bursts that the classical search had already found, so 12 were pulses that only the CNN had found. The authors report that the 16 candidates that they missed were faint pulses whose astrophysical nature is unclear. The rest of the observation was not searched again."", ""url"": ""https://arxiv.org/html/2307.02303v2""}]","{""kind"": ""Challenges""}" AIBD-0243,2023-02-14,event,day,reported,Camera-based AI feedback on 100 Spin e-scooters in Santa Monica cuts time ridden on sidewalks by 22% in a randomized field test,"Camera-based computer vision from an unnamed maker detected sidewalk riding on Spin e-scooters in Santa Monica. In a randomized test, its feedback cut the time ridden on sidewalks by 22%.",Camera AI on Spin e-scooters cuts time ridden on sidewalks 22%,The feedback group spent 22% less time and 20% less distance on sidewalks,,reported,Independently reported,randomized trial,moderate,aggregate result,false,"Camera-based computer-vision riding-surface detection with in-app notifications, audible beeps and speed reduction on Spin shared e-scooters (maker not named in the paper)",,Not named in the paper,,"Spin (shared e-scooter operator), with University of Washington researchers",United States,"Santa Monica, United States",United States,Micromobility safety,Shared e-scooters / pedestrian conflict,Safety and rescue,Images and video,Drivers and travelers,11,High income,physical_health,Hazard reduction,2,,3,,3,,,societal; organizational; personal,2,2,the no-feedback group (50 scooters with feedback mechanisms disabled at random),"0.22 (mean fraction of trip time on sidewalks, no-feedback group, Table 1)","0.17 (mean fraction of trip time on sidewalks, feedback group, Table 1)",measured,withdrawn,2024-11-20,shown,https://findingspress.org/article/125892-ai-driven-feedback-systems-reduce-sidewalk-riding-on-shared-e-scooters,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Follow-up"", ""value"": ""The study's own limitations section says that Spin withdrew all camera-equipped scooters from service after they were targeted by thieves. The withdrawal is dated only by the paper's publication."", ""url"": ""https://findingspress.org/article/125892-ai-driven-feedback-systems-reduce-sidewalk-riding-on-shared-e-scooters""}]","{""kind"": ""Withdrawals""}" AIBD-0346,2023-02-01,event,month,reported,Responders in Türkiye used machine-learning maps of damage from xView2 after the earthquakes of 6 February 2023,"xView2 is a machine-learning model that grades building damage in satellite images. After the 2023 Türkiye earthquakes, it gave responders maps that UN rescue teams used in Adıyaman.",Rescue workers in Türkiye find damaged areas with xView2 maps,would typically take weeks and now takes hours or minutes,,reported,Independently reported,no comparison,very low,protective response,false,"xView2 building damage assessment models (semantic segmentation on satellite imagery, trained on the xBD dataset)",,US Defense Innovation Unit (DIU) with Carnegie Mellon University Software Engineering Institute; research partners Microsoft AI for Good and UC Berkeley,United States,"UN INSARAG (search and rescue ground teams in Adıyaman); AFAD, World Bank, IFRC and WFP (per the developer's account); Microsoft AI for Good delivered estimates to AFAD (per Microsoft)",Switzerland,Adıyaman and four cities in southeast Türkiye including Kahramanmaraş (Marash),Turkey,Disaster response / post-earthquake damage assessment,Humanitarian and civil-protection organizations,Humanitarian,Images and video,Public servants,11,Upper middle income,operational_improvement,"Faster data management, decision-making",2,2,1,,3,,,personal; organizational; societal,5,2,conventional damage assessment,weeks,hours or minutes,assumed,challenged,2025-10-27,shown,https://www.technologyreview.com/2023/02/20/1068824/ai-actually-helpful-disaster-response-turkey-syria-earthquake/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""Researchers from Stanford and Direct Relief tested the Microsoft model built on xBD against reference damage data for hospitals and dialysis centers in Antakya, Kahramanmaraş and Iskenderun. They report that it found 42.86% of damaged facilities, with 93.55% specificity. They conclude that such models cannot yet replace ground surveys."", ""url"": ""https://pmc.ncbi.nlm.nih.gov/articles/PMC12558478/""}]","{""kind"": ""Challenges""}" -AIBD-0003,2022-09-01,event,month,reported,"Brain implant at the University of California, San Francisco, decodes a woman's attempted speech at 78 words per minute 17 years after a stroke, before its removal in February 2024 for an unrelated reason","A brain implant at the University of California, San Francisco, decoded Ann Johnson's attempts to speak at a median 78 words per minute. It drove a speaking avatar.",Brain implant lets a woman speak via an avatar 17 years after a stroke,decoded text at a median 78 words per minute,,reported,Independently reported,,very low,benefit created,false,UCSF multimodal speech neuroprosthesis (BRAVO trial),"253-channel high-density ECoG array; bidirectional RNN decoders with CTC loss producing text, synthesized speech and avatar gestures; 1,024-word vocabulary",UCSF Chang Lab with UC Berkeley,United States,UCSF Medical Center,United States,"San Francisco, United States",United States,Healthcare / assistive communication,Post-stroke anarthria,Accessibility,Prediction from data; Speech and sound,People with disabilities; Patients,3; 10,High income,physical_health; accessibility,,2,1,3,,5,5,,personal,6,2,"the commercial device that she usually used, which tracks the movement of her head",about 14.2 words per minute with her assistive device,"a median 78.3 words per minute of decoded text, with a median word error rate of 25% over a 1,024-word vocabulary",measured,withdrawn,2024-02-01,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC10826467/,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Follow-up"", ""value"": ""Johnson took part in the study of streaming speech synthesis in 2025, but its data predate the removal of her implant. Berkeley News reported that the implant was removed in February 2024 for a reason unrelated to the trial. It also reported that she still writes to the team with the assistive technology that she uses now."", ""url"": ""https://news.berkeley.edu/2025/08/13/a-stroke-stole-her-ability-to-speak-eighteen-years-later-scientists-used-ai-to-bring-it-back/""}]","{""kind"": ""Withdrawals""}" +AIBD-0003,2022-09-01,event,month,reported,"Brain implant at the University of California, San Francisco, decodes a woman's attempted speech at 78 words per minute 17 years after a stroke, before its removal in February 2024 for an unrelated reason","A brain implant at the University of California, San Francisco, decoded Ann Johnson's attempts to speak at a median 78 words per minute. It drove a speaking avatar.",Brain implant lets a woman speak via an avatar 17 years after a stroke,decoded text at a median 78 words per minute,,reported,Independently reported,before and after,very low,benefit created,false,UCSF multimodal speech neuroprosthesis (BRAVO trial),"253-channel high-density ECoG array; bidirectional RNN decoders with CTC loss producing text, synthesized speech and avatar gestures; 1,024-word vocabulary",UCSF Chang Lab with UC Berkeley,United States,UCSF Medical Center,United States,"San Francisco, United States",United States,Healthcare / assistive communication,Post-stroke anarthria,Accessibility,Prediction from data; Speech and sound,People with disabilities; Patients,3; 10,High income,physical_health; accessibility,,2,1,3,,5,5,,personal,6,2,"the commercial device that she usually used, which tracks the movement of her head",about 14.2 words per minute with her assistive device,"a median 78.3 words per minute of decoded text, with a median word error rate of 25% over a 1,024-word vocabulary",measured,withdrawn,2024-02-01,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC10826467/,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Follow-up"", ""value"": ""Johnson took part in the study of streaming speech synthesis in 2025, but its data predate the removal of her implant. Berkeley News reported that the implant was removed in February 2024 for a reason unrelated to the trial. It also reported that she still writes to the team with the assistive technology that she uses now."", ""url"": ""https://news.berkeley.edu/2025/08/13/a-stroke-stole-her-ability-to-speak-eighteen-years-later-scientists-used-ai-to-bring-it-back/""}]","{""kind"": ""Withdrawals""}" AIBD-0031,2022-06-01,event,month,reported,"Randomized evaluation of volunteers relaying AI-based flood warnings in Bihar, India (2022–2025 monsoons)","In a trial in Bihar, volunteers relayed flood alerts from Google and the Central Water Commission. Under severe floods, their villages' homes were 12.7 percentage points less likely to flood.",Flood alerts relayed by volunteers cut home flooding in Bihar,12.7 percentage points less likely to have their own home flooded,,reported,Independently reported,randomized trial,low,aggregate result,false,"Google flood forecasting with Android push alerts (Google Public Alerts), plus community dissemination by Yuganter volunteers",,Google,United States,Central Water Commission of India with Google; last-mile dissemination by Yuganter (NGO) under the Jagnani-Pande evaluation,India,"Bihar (twelve flood-prone districts), India",India,Disaster risk reduction / flood early warning,Rural households in riverine flood zones,Humanitarian,Prediction from data,People at risk of disaster,13,Lower middle income,environmental_impact,Natural disaster prediction,2,5,4,4,4,2,0.450,personal; societal,4,1,control village councils that received the same CWC and Google forecasts and alerts but had no community volunteers,"control households under severe flooding, whose chance of home flooding was 23.7 percentage points higher than in a season with no flood",treated households under severe flooding were 12.7 percentage points less likely to have their own home flooded,measured,challenged,2026-09-01,shown,https://arxiv.org/abs/2111.02780,"INR 6,500",less spent on illness treatment per household over the flood season,False,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""In the September 2026 version, which covers four seasons from 2022 to 2025, the same authors report no treatment effect on physical health. The 16.6 percent drop in illness symptoms and the lower medical spending from the April 2024 version are not repeated. They report gains in home flooding, food security and borrowing instead."", ""url"": ""https://www.dropbox.com/scl/fi/gpwdwa4hkm9e2e81c2ne5/floods_EWS.pdf?rlkey=3n4l804arm1zms5y0pt0ac5sp&raw=1""}]","{""kind"": ""Challenges""}" AIBD-0369,2022-03-10,event,day,reported,Westchester County police use Rekor's analysis of plate-reader patterns to stop a man in Scarsdale. Officers find crack cocaine. He is convicted,Rekor Scout software flagged a man's trips past Westchester County's 480 plate-reader cameras. Officers who stopped him in Scarsdale on 10 March 2022 found crack cocaine. They found a pistol.,Police find drugs in Scarsdale after Rekor flags a driver's trips,,,reported,Independently reported,no comparison,very low,protective response,false,Rekor license-plate-reader pattern analysis (Rekor Public Safety Network data) used by the Westchester County Police Department,Rekor Scout (Interdiction Analysis report),Rekor,United States,Westchester County Police Department,United States,"Scarsdale, United States",United States,Policing / drug trafficking,Government,Public sector,Images and video; Prediction from data,Public servants,16,High income,physical_health,Crime reduction or prevention,2,,4,,3,1,,societal,5,0,,,,,persisted,2025-01-17,not shown,https://static.foxnews.com/foxnews.com/content/uploads/2023/07/david-zayas-march-10-court-filing-defense.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""United States v. Zayas, 22 Cr. 178 (NSR), defense motion filed 10 March 2023, pp. 1–6 and footnote 8. Direct litigation account of the license-plate search and disputed analytic report. Statements are the defense’s account. They are not judicial findings."", ""url"": ""https://static.foxnews.com/foxnews.com/content/uploads/2023/07/david-zayas-march-10-court-filing-defense.pdf""}, {""label"": ""Court document"", ""value"": ""U.S. District Court for the Southern District of New York, United States v. Zayas, 7:22-cr-00178 (NSR), Judgment in a Criminal Case. Signed by Judge Nelson S. Román on 30 April 2024 (docket text in the RECAP archive). The court has no stake in the system. The judgment records the guilty plea, the 48-month sentence and the forfeiture of $34,000. The analysis from the plate readers does not appear in it."", ""url"": ""https://www.courtlistener.com/?q=docket_id%3A63178722&type=r&order_by=entry_date_filed+desc""}]","{""kind"": ""Official documents""}" AIBD-0167,2018-08-01,event,day,reported,"IdentiFlight stops wind turbines when it detects eagles at a Wyoming wind farm, where disputed estimates put the cut in eagle deaths at 50% to 85%","Researchers estimated an 82% (75%–89%) relative cut in eagle deaths at a Wyoming wind farm after IdentiFlight began curtailing turbines. Two federal statisticians later put it at 50% (−159%, 89%).",How many eagles IdentiFlight's turbine stops save is disputed,82% (75%–89%) relative cut in eagle deaths at a Wyoming wind farm,,reported,Independently reported,controlled comparison,very low,aggregate result,false,IdentiFlight automated eagle detection and turbine curtailment (camera system with machine-learning classification),,IdentiFlight (Boulder Imaging),United States,"Top of the World Windpower Facility, Wyoming",United States,"Wyoming, United States",United States,Energy,Wind power wildlife mitigation,Energy and infrastructure,Images and video,,15,High income,environmental_impact; animal_health,Environmental impacts,2,,5,,4,,,societal; organizational,7,3,"a control wind facility 15 km away without automated curtailment, before and after the installation",increasing at the control site by 113% (51%–218%),the number of fatalities at the treatment site declined by 63% (95% CI = 59%–66%),measured,persisted,,shown,https://doi.org/10.1371/journal.pone.0278754,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Fish and Wildlife Service, Permits for Incidental Take of Eagles and Eagle Nests (final rule), Federal Register 89 FR 9920, 12 February 2024. The Service regulates eagle take and has no stake in IdentiFlight. It allows permittees to use the system but finds that the efficacy of such technologies has not been studied enough to require them."", ""url"": ""https://www.federalregister.gov/documents/2024/02/12/2024-02182/permits-for-incidental-take-of-eagles-and-eagle-nests""}]","{""kind"": ""Official documents""}" @@ -409,23 +409,23 @@ AIBD-0144,2019-01-01,event,year,reported,"Blackpool Council reports cheaper poth AIBD-0922,2026-06-01,source publication,month,reported,"UNDP's evaluation office reports that at three Zimbabwean irrigation schemes, drones with AI monitored crops, enabling earlier responses to threats",UNDP's evaluation office reported that drones with AI monitored crop health at three Zimbabwean irrigation schemes. This sped up responses to threats of crop loss. The responses were more precise.,Drones with AI watch crop health at three Zimbabwean irrigation schemes,,,reported,Independently reported,no comparison,very low,protective response,false,AI and machine-learning crop health monitoring with drone technology (product not named),,Not named in the evaluation,,"UNDP Zimbabwe, through the resilience-building project funded by the Green Climate Fund and run with the Government of Zimbabwe",Zimbabwe,"Midlo (Matabeleland South), Pikinini/Jawanda and Bindamombe (Masvingo) irrigation schemes, Zimbabwe",Zimbabwe,Agriculture / crop protection,,Agriculture and food,Images and video,Farmers,2,Lower middle income,environmental_impact,Environmental impacts,,,,4,3,,,personal,2,0,,,,,unknown,,shown,https://erc.undp.org/evaluation/documents/download/25595,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Sanhokwe, H. and Diouf, A. Interim evaluation of the 'Building Climate Resilience of Vulnerable Agricultural Livelihoods in Southern Zimbabwe' project: Final report. Interim evaluation under the Green Climate Fund's guidelines for evaluations led by accredited entities, published in the UNDP Evaluation Resource Center, 6 September 2024, p. 48."", ""url"": ""https://erc.undp.org/evaluation/documents/download/23866""}]","{""kind"": ""Official documents""}" AIBD-0245,2026-04-24,event,day,reported,"After a Tesla on Autopilot stops in the middle lane of I-75 with its drunk driver asleep, police arrest her for driving under the influence",Troopers said that the eye detection of a Tesla on Autopilot halted the car on I-75 after its intoxicated driver fell asleep. No vehicle data have been published to confirm it.,Tesla halts on I-75 when its drunk driver falls asleep at the wheel,,,reported,Independently reported,no comparison,very low,protective response,false,"Tesla Autopilot, according to the reported account; the vehicle model and active configuration are unverified.",,Tesla,United States,Private owner,,"Sarasota, United States",United States,Road vehicle crash avoidance,Consumer driver assistance / driver monitoring,Safety and rescue,Machines that act; Images and video,Drivers and travelers,3,High income,physical_health,,,1,1,,5,1,,personal; societal; organizational,4,2,,,,,unknown,,shown,https://en.wikipedia.org/wiki/Tesla_Autopilot,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""NHTSA Office of Defects Investigation, ODI Resume, Engineering Analysis EA22-002: Autopilot System Driver Controls, closed 25 April 2024. Federal regulators found that Autopilot's controls for driver engagement were too weak for what the system allowed. Tesla filed a recall of 2,031,220 cars on 12 December 2023 after NHTSA reviewed these findings with it."", ""url"": ""https://static.nhtsa.gov/odi/inv/2022/INCLA-EA22002-14498.pdf""}]","{""kind"": ""Official documents""}" AIBD-0238,2025-08-08,source publication,day,tested,DARPA's AIxCC final: Autonomous systems find 54 of 63 synthetic vulnerabilities. They patch 43. They also find 18 real bugs in open-source code,"At DEF CON 33 on 8 August 2025, DARPA reported that seven teams' cyber reasoning systems found 54 synthetic vulnerabilities, patched 43, and found 18 real bugs for disclosure.","AIxCC systems find 54 of 63 synthetic vulnerabilities, patching 43","found 54 synthetic vulnerabilities, patched 43, and found 18 real bugs for disclosure",,reported,Independently reported,tested against a standard,very low,aggregate result,false,"AIxCC finalist cyber reasoning systems (Team Atlanta, Trail of Bits, Theori and four others)",,"Seven finalist teams (Team Atlanta et al.), using LLM credits from Anthropic, Google and OpenAI",,"DARPA (with ARPA-H), AIxCC Final Competition",United States,"Las Vegas, United States",United States,Cybersecurity / automated vulnerability discovery and patching,Government,Public sector,Text and language,,9,High income,privacy_protection,,,,,,,,,societal; organizational,6,0,the same competition's semifinal round (2024),37% at semifinals,teams identified 86% of the competition's synthetic vulnerabilities,measured,challenged,2026-09-03,shown,https://www.darpa.mil/news/2025/aixcc-results,$152,"average cost per competition task for the systems' bug reports and patches, which DARPA calls a fraction of the cost of traditional methods",False,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""University of Maryland researchers tested the top three AIxCC systems on a new benchmark in September 2026. They found that crash-only checks overstate patching success. They also found that the AIxCC systems solved fewer tasks than general-purpose coding agents on the same model. In March 2026 Team Atlanta reported that the seven released systems were largely unusable outside their teams and that its ported system found 10 new bugs."", ""url"": ""https://arxiv.org/html/2609.04075v1""}]","{""kind"": ""Challenges""}" -AIBD-0325,2025-05-24,source publication,day,unclear,A hospital's account conflicts with a relative’s claim that ChatGPT helped save a teenager from poisoning,A relative credited ChatGPT's advice to drink paraffin oil with saving a teenager from a grain pill. The medical dean at Assiut University said that the teenager showed no signs of poisoning.,Hospital disputes that ChatGPT helped save a teenager from poisoning,,,reported,Independently reported,,not rated,protective response,true,"ChatGPT, according to the patient's cousin. He described the model as an advanced AI model and a limited-use trial version of GPT-5, but that identification is unverified. OpenAI publicly introduced GPT-5 on 7 August 2025, after the May report.","GPT-5 trial version, per the cousin's post; unverified",OpenAI,United States,"The patient's cousin, as a consumer user",,"Abnoub and Assiut University Hospital, Assiut Governorate, Egypt",Egypt,Healthcare: emergency toxicology,Consumer chatbot used for first-aid advice,Health,Text and language,Children; Patients,3,Lower middle income,physical_health,,,,,,,,,personal; societal,7,0,"the doctors' prognosis on arrival at the poison unit, by the cousin's account","no treatment protocol, and the case would die","discharged after 24 hours, by the cousin's account",assumed,reversed,2025-05-25,shown,https://web.archive.org/web/2026id_/https://openai.com/index/helping-people-when-they-need-it-most/,,,,[],Querist,The benefit was later reversed.,"[{""label"": ""Follow-up"", ""value"": ""The next day the dean of Assiut University's Faculty of Medicine told Masrawy that the youth showed no clinical signs of phosphine poisoning and had categorically denied swallowing the pill. On 26 May a toxicologist told Al Jazeera Mubasher that the claim that paraffin oil rescues such a case has no scientific basis."", ""url"": ""https://www.masrawy.com/news/news_regions/details/2025/5/25/2793185/""}]","{""kind"": ""Challenges""}" -AIBD-0326,2025-05-07,event,day,unclear,"Rafael's X-Guard towed decoy, labeled “AI-powered”, may have diverted Pakistani missiles from Indian Rafales in Operation Sindoor (7–10 May 2025). India's defense chief reports Indian aircraft losses. So does France's air force chief","Commentators credit Rafael's X-Guard towed decoy, labeled “AI-powered”, with diverting Pakistani missiles from Indian Rafales. No official has confirmed its use. India admitted jet losses.",X-Guard decoy may have diverted Pakistani missiles from Indian jets,,,reported,Independently reported,,not rated,protective response,false,X-Guard fiber-optic towed decoy from Rafael Advanced Defense Systems. The manufacturer describes its integration with an aircraft’s electronic warfare system but does not identify an AI model or confirm use in Operation Sindoor.,,Rafael Advanced Defense Systems (Israel); several outlets misname it 'Rafale Advanced Defense Systems',Israel,Indian Air Force (claimed by commentary; unconfirmed by any official source),India,"India-Pakistan border airspace, Operation Sindoor, 7-10 May 2025",India | Pakistan,Defense: electronic warfare,Airborne towed radar decoys,Defense,,,,Lower middle income,physical_health,,,,,,,,,personal; organizational,10,0,,,,,reversed,2025-05-31,not shown,https://www.rafael.co.il/wp-content/uploads/2024/03/X-Guard-Brochure-English-03-2024.pdf,,,,[],Querist,The benefit was later reversed.,"[{""label"": ""Follow-up"", ""value"": ""India's Chief of Defence Staff admitted on 31 May 2025 that an unspecified number of Indian jets were shot down in the May conflict. That contradicts the outlets' claim that the decoy kept missiles from hitting real aircraft."", ""url"": ""https://www.aljazeera.com/news/2025/6/1/india-general-admits-jet-losses-in-clash-with-pakistan-heres-what-he-said""}]","{""kind"": ""Challenges""}" -AIBD-0150,2025-01-01,event,year,unclear,"NBC News reported that a Long Beach, California, tenant who lost an eviction jury trial with a lawyer overturned the eviction on appeal with ChatGPT and Perplexity, though court records show that her later petitions were denied in November 2025","NBC News reported that Lynn White used ChatGPT and Perplexity to overturn her eviction on appeal. According to court records, her later petitions in the case were denied in November 2025.",Tenant may have overturned her eviction with ChatGPT and Perplexity,"avoided roughly $55,000 in penalties and over $18,000 in overdue rent",,reported,Independently reported,,not rated,benefit created,false,"ChatGPT (paid plan, $20 a month) and Perplexity Pro ($20 a month)","Not reported; NBC News says White moved from the free ChatGPT tier to the $20-a-month plan and used Perplexity Pro, without naming model versions",OpenAI (ChatGPT) and Perplexity AI (Perplexity),United States,,,"Long Beach, United States",United States,Access to justice,Self-represented tenants in eviction proceedings,Public sector,Text and language,Consumers,16,High income,liberty_protection,,,,,,,,,personal,4,0,"representation by a lawyer from a local tenant advocacy network, with whom she lost the jury trial",she worked with a lawyer from a local tenant advocacy network but lost in a jury trial,"White managed to overturn her eviction notice and avoid roughly $55,000 in penalties and more than $18,000 in overdue rent",assumed,litigated,2025-11-14,shown,https://web.archive.org/web/2023id_/https://openai.com/blog/chatgpt,"roughly $55,000","penalties avoided, plus more than $18,000 in overdue rent, by overturning the eviction on appeal",False,[],Querist,The use of the system was taken to court.,"[{""label"": ""Follow-up"", ""value"": ""On 5 November 2025 White, representing herself, petitioned the Court of Appeal, Second District, for a writ of mandate and a stay. The petition was against the Los Angeles Superior Court and Judge Ronald F. Frank of the Long Beach courthouse. Best Trailer Park, LLC was the real party in interest. The trial court cases are the eviction case 23LBUD01928 and the appellate division case 24APLC00165. The court denied the petition on 14 November 2025 and rejected a motion to vacate the denial on 17 November. The California Supreme Court allowed a late petition for review with a stay request and denied both on 26 November 2025 (S294091)."", ""url"": ""https://appellatecases.courtinfo.ca.gov/search/searchResults.cfm?dist=2&search=number&query_caseNumber=B350202""}]","{""kind"": ""Legal proceedings""}" -AIBD-0338,2023-04-01,event,month,intended,"IRS uses AI models to pick returns of large partnerships for audit, but no examination that a model selected had closed by May 2025","The IRS Large Partnership Compliance Model, an AI model, flagged 150 returns of large partnerships for review in April 2023. No examination that it selected had closed by May 2025.",IRS applies AI to select large partnerships for examination,,,reported,Independently reported,,not rated,program,false,"IRS AI and machine-learning case-selection models (two partnership prioritization models; a refundable-credit audit selection model in pilot; machine learning for partnership, income and international tax risk)",TY 2021 LPC model (Large Partnership Compliance return selection model),Internal Revenue Service (in-house data scientists with tax-enforcement experts),United States,Internal Revenue Service,United States,United States (national),United States,Public administration: tax enforcement,National revenue agency,Finance,Prediction from data,Public servants,16,High income,privacy_protection; operational_improvement,Fraud detection,,,,,,,,personal; organizational; societal,5,0,large partnership audits selected before the Large Partnership Compliance Model,more than 80 percent of large partnership audits closed with no change to the return,not yet measured; TIGTA reported in May 2025 that no examination selected by the model had closed,measured,persisted,2025-05-19,shown,https://www.oversight.gov/sites/default/files/documents/reports/2025-05/2025308022fr.pdf,approximately $26.1 million,"additional tax assessments yearly that IRS management estimates from reassigning 14 classifiers to examinations after the Form 1040 return classification AI model automated classification (an SB/SE individual-return model, not the partnership models)",True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Treasury Inspector General for Tax Administration (2025). The IRS Could Leverage Examination Results in Artificial Intelligence Examination Case Selection Models and Improve Processes to Evaluate Performance. Report 2025-308-022, 19 May 2025."", ""url"": ""https://www.oversight.gov/sites/default/files/documents/reports/2025-05/2025308022fr.pdf""}, {""label"": ""Audit report"", ""value"": ""U.S. Government Accountability Office, Tax Enforcement: IRS Audit Processes Can Be Strengthened to Address a Growing Number of Large, Complex Partnerships, GAO-23-106020, 27 July 2023. GAO has no stake in the IRS models. Its auditors report that the IRS first used the Large Partnership Compliance Model in April 2023. They also report that the partnership models were built without representative samples and with untested assumptions."", ""url"": ""https://www.gao.gov/assets/gao-23-106020.pdf""}, {""label"": ""Audit report"", ""value"": ""Treasury Inspector General for Tax Administration (TIGTA). The IRS Has Yet to Develop a Successful Strategy for Examining Large Partnership Returns (Report Number 2026-308-011). 'Preliminary results of the LPC Program', report p. 10, Figure 5"", ""url"": ""https://www.oversight.gov/sites/default/files/documents/reports/2026-03/2026308011fr.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0325,2025-05-24,source publication,day,unclear,A hospital's account conflicts with a relative’s claim that ChatGPT helped save a teenager from poisoning,A relative credited ChatGPT's advice to drink paraffin oil with saving a teenager from a grain pill. The medical dean at Assiut University said that the teenager showed no signs of poisoning.,Hospital disputes that ChatGPT helped save a teenager from poisoning,,,reported,Independently reported,,very low,protective response,true,"ChatGPT, according to the patient's cousin. He described the model as an advanced AI model and a limited-use trial version of GPT-5, but that identification is unverified. OpenAI publicly introduced GPT-5 on 7 August 2025, after the May report.","GPT-5 trial version, per the cousin's post; unverified",OpenAI,United States,"The patient's cousin, as a consumer user",,"Abnoub and Assiut University Hospital, Assiut Governorate, Egypt",Egypt,Healthcare: emergency toxicology,Consumer chatbot used for first-aid advice,Health,Text and language,Children; Patients,3,Lower middle income,physical_health,,,,,,,,,personal; societal,7,0,"the doctors' prognosis on arrival at the poison unit, by the cousin's account","no treatment protocol, and the case would die","discharged after 24 hours, by the cousin's account",assumed,reversed,2025-05-25,shown,https://web.archive.org/web/2026id_/https://openai.com/index/helping-people-when-they-need-it-most/,,,,[],Querist,The benefit was later reversed.,"[{""label"": ""Follow-up"", ""value"": ""The next day the dean of Assiut University's Faculty of Medicine told Masrawy that the youth showed no clinical signs of phosphine poisoning and had categorically denied swallowing the pill. On 26 May a toxicologist told Al Jazeera Mubasher that the claim that paraffin oil rescues such a case has no scientific basis."", ""url"": ""https://www.masrawy.com/news/news_regions/details/2025/5/25/2793185/""}]","{""kind"": ""Challenges""}" +AIBD-0326,2025-05-07,event,day,unclear,"Rafael's X-Guard towed decoy, labeled “AI-powered”, may have diverted Pakistani missiles from Indian Rafales in Operation Sindoor (7–10 May 2025). India's defense chief reports Indian aircraft losses. So does France's air force chief","Commentators credit Rafael's X-Guard towed decoy, labeled “AI-powered”, with diverting Pakistani missiles from Indian Rafales. No official has confirmed its use. India admitted jet losses.",X-Guard decoy may have diverted Pakistani missiles from Indian jets,,,reported,Independently reported,,very low,protective response,false,X-Guard fiber-optic towed decoy from Rafael Advanced Defense Systems. The manufacturer describes its integration with an aircraft’s electronic warfare system but does not identify an AI model or confirm use in Operation Sindoor.,,Rafael Advanced Defense Systems (Israel); several outlets misname it 'Rafale Advanced Defense Systems',Israel,Indian Air Force (claimed by commentary; unconfirmed by any official source),India,"India-Pakistan border airspace, Operation Sindoor, 7-10 May 2025",India | Pakistan,Defense: electronic warfare,Airborne towed radar decoys,Defense,,,,Lower middle income,physical_health,,,,,,,,,personal; organizational,10,0,,,,,reversed,2025-05-31,not shown,https://www.rafael.co.il/wp-content/uploads/2024/03/X-Guard-Brochure-English-03-2024.pdf,,,,[],Querist,The benefit was later reversed.,"[{""label"": ""Follow-up"", ""value"": ""India's Chief of Defence Staff admitted on 31 May 2025 that an unspecified number of Indian jets were shot down in the May conflict. That contradicts the outlets' claim that the decoy kept missiles from hitting real aircraft."", ""url"": ""https://www.aljazeera.com/news/2025/6/1/india-general-admits-jet-losses-in-clash-with-pakistan-heres-what-he-said""}]","{""kind"": ""Challenges""}" +AIBD-0150,2025-01-01,event,year,unclear,"NBC News reported that a Long Beach, California, tenant who lost an eviction jury trial with a lawyer overturned the eviction on appeal with ChatGPT and Perplexity, though court records show that her later petitions were denied in November 2025","NBC News reported that Lynn White used ChatGPT and Perplexity to overturn her eviction on appeal. According to court records, her later petitions in the case were denied in November 2025.",Tenant may have overturned her eviction with ChatGPT and Perplexity,"avoided roughly $55,000 in penalties and over $18,000 in overdue rent",,reported,Independently reported,,very low,benefit created,false,"ChatGPT (paid plan, $20 a month) and Perplexity Pro ($20 a month)","Not reported; NBC News says White moved from the free ChatGPT tier to the $20-a-month plan and used Perplexity Pro, without naming model versions",OpenAI (ChatGPT) and Perplexity AI (Perplexity),United States,,,"Long Beach, United States",United States,Access to justice,Self-represented tenants in eviction proceedings,Public sector,Text and language,Consumers,16,High income,liberty_protection,,,,,,,,,personal,4,0,"representation by a lawyer from a local tenant advocacy network, with whom she lost the jury trial",she worked with a lawyer from a local tenant advocacy network but lost in a jury trial,"White managed to overturn her eviction notice and avoid roughly $55,000 in penalties and more than $18,000 in overdue rent",assumed,litigated,2025-11-14,shown,https://web.archive.org/web/2023id_/https://openai.com/blog/chatgpt,"roughly $55,000","penalties avoided, plus more than $18,000 in overdue rent, by overturning the eviction on appeal",False,[],Querist,The use of the system was taken to court.,"[{""label"": ""Follow-up"", ""value"": ""On 5 November 2025 White, representing herself, petitioned the Court of Appeal, Second District, for a writ of mandate and a stay. The petition was against the Los Angeles Superior Court and Judge Ronald F. Frank of the Long Beach courthouse. Best Trailer Park, LLC was the real party in interest. The trial court cases are the eviction case 23LBUD01928 and the appellate division case 24APLC00165. The court denied the petition on 14 November 2025 and rejected a motion to vacate the denial on 17 November. The California Supreme Court allowed a late petition for review with a stay request and denied both on 26 November 2025 (S294091)."", ""url"": ""https://appellatecases.courtinfo.ca.gov/search/searchResults.cfm?dist=2&search=number&query_caseNumber=B350202""}]","{""kind"": ""Legal proceedings""}" +AIBD-0338,2023-04-01,event,month,intended,"IRS uses AI models to pick returns of large partnerships for audit, but no examination that a model selected had closed by May 2025","The IRS Large Partnership Compliance Model, an AI model, flagged 150 returns of large partnerships for review in April 2023. No examination that it selected had closed by May 2025.",IRS applies AI to select large partnerships for examination,,,reported,Independently reported,,very low,program,false,"IRS AI and machine-learning case-selection models (two partnership prioritization models; a refundable-credit audit selection model in pilot; machine learning for partnership, income and international tax risk)",TY 2021 LPC model (Large Partnership Compliance return selection model),Internal Revenue Service (in-house data scientists with tax-enforcement experts),United States,Internal Revenue Service,United States,United States (national),United States,Public administration: tax enforcement,National revenue agency,Finance,Prediction from data,Public servants,16,High income,privacy_protection; operational_improvement,Fraud detection,,,,,,,,personal; organizational; societal,5,0,large partnership audits selected before the Large Partnership Compliance Model,more than 80 percent of large partnership audits closed with no change to the return,not yet measured; TIGTA reported in May 2025 that no examination selected by the model had closed,measured,persisted,2025-05-19,shown,https://www.oversight.gov/sites/default/files/documents/reports/2025-05/2025308022fr.pdf,approximately $26.1 million,"additional tax assessments yearly that IRS management estimates from reassigning 14 classifiers to examinations after the Form 1040 return classification AI model automated classification (an SB/SE individual-return model, not the partnership models)",True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Treasury Inspector General for Tax Administration (2025). The IRS Could Leverage Examination Results in Artificial Intelligence Examination Case Selection Models and Improve Processes to Evaluate Performance. Report 2025-308-022, 19 May 2025."", ""url"": ""https://www.oversight.gov/sites/default/files/documents/reports/2025-05/2025308022fr.pdf""}, {""label"": ""Audit report"", ""value"": ""U.S. Government Accountability Office, Tax Enforcement: IRS Audit Processes Can Be Strengthened to Address a Growing Number of Large, Complex Partnerships, GAO-23-106020, 27 July 2023. GAO has no stake in the IRS models. Its auditors report that the IRS first used the Large Partnership Compliance Model in April 2023. They also report that the partnership models were built without representative samples and with untested assumptions."", ""url"": ""https://www.gao.gov/assets/gao-23-106020.pdf""}, {""label"": ""Audit report"", ""value"": ""Treasury Inspector General for Tax Administration (TIGTA). The IRS Has Yet to Develop a Successful Strategy for Examining Large Partnership Returns (Report Number 2026-308-011). 'Preliminary results of the LPC Program', report p. 10, Figure 5"", ""url"": ""https://www.oversight.gov/sites/default/files/documents/reports/2026-03/2026308011fr.pdf""}]","{""kind"": ""Official documents""}" AIBD-0111,2022-01-05,event,day,reported,Google's Android Earthquake Alerts reached phones in Alanya seconds before a quake of magnitude 5.1 in the Mediterranean on 5 January 2022,"Google's Android Earthquake Alerts System sent Android users near Alanya, Turkey, a warning seconds before a magnitude 5.1 offshore quake on 5 January 2022.",Android alerts reach phones in Alanya seconds before an offshore quake,Hürriyet put the lead at about 10 seconds,,reported,Independently reported,no comparison,very low,protective response,false,Android Earthquake Alerts System (crowdsourced phone-accelerometer detection; alerts for magnitude 4.5 and above),,Google,United States,Google,United States,"Alanya, Turkey",Turkey,Disaster risk reduction / earthquake early warning,Consumer smartphones (Android),Humanitarian,Prediction from data,People at risk of disaster,11,Upper middle income,environmental_impact,Natural disaster prediction,,,1,,4,1,,personal,16,2,"Feeling the shaking with no advance notice, as before the system reached Turkey",No warning before the shaking,"A phone alert about 10 seconds before the shaking, by Hürriyet's account; other outlets gave up to one minute",assumed,challenged,2023-07-27,not shown,https://blog.google/products/android/earthquake-detection-and-alerts/,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""BBC Newsnight found no one in three Turkish cities who had received a warning from the same system before the main shock of the deadly earthquake of 6 February 2023. Google's product lead told the BBC that he was confident that the system fired and sent alerts. Google did not give the BBC evidence that the alerts were widely received."", ""url"": ""https://www.bbc.com/news/technology-66316462""}]","{""kind"": ""Challenges""}" AIBD-0203,2021-02-01,event,month,mixed,FDA uses AI to choose where to concentrate food inspections. An imported-seafood pilot ran for years,"FDA reports that its model for imported seafood guided sampling at 328 U.S. ports. It says that its Elsa tool helped find inspection targets. It reports no outcome, such as illnesses prevented.",FDA uses AI to choose where to inspect food,,,reported,Independently reported,no comparison,very low,program,false,FDA AI inspection-targeting tools (imported seafood pilot; Elsa generative AI),,U.S. Food and Drug Administration (details not in the cited source),United States,U.S. Food and Drug Administration,United States,United States (national),United States,Food safety regulation,Import and facility inspection targeting,Agriculture and food,Prediction from data; Text and language,Public servants,3,High income,operational_improvement,"Faster data management, decision-making",,1,3,,2,5,,organizational; societal,4,0,,,,,challenged,2025-07-23,shown,https://www.fda.gov/food/hfp-constituent-updates/fda-moves-third-phase-artificial-intelligence-imported-seafood-pilot-program,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""FDA launched Elsa agency-wide on 2 June 2025 and said that it was already helping to identify high-priority inspection targets. On 23 July 2025 CNN reported that FDA staff said that Elsa invents studies, and Makary told CNN that most scientists used it for organizing work."", ""url"": ""https://www.cnn.com/2025/07/23/politics/fda-ai-elsa-drug-regulation-makary""}]","{""kind"": ""Challenges""}" -AIBD-0334,2020-07-27,source publication,day,intended,"Colombia's Constitutional Court presents PretorIA (July 2020), an AI system that classifies tutela rulings for selection. It was in user testing in April 2021. It was still in use in 2024, with no published measure of an outcome","Colombia's Constitutional Court presented PretorIA on 27 July 2020 to classify tutela rulings for selection. It was still in use in 2024, with no published measure of outcomes.",Colombia's Constitutional Court adopts PretorIA to sort tutela rulings,,,reported,Independently reported,,not rated,program,false,"PretorIA (Corte Constitucional tutela classification and search system, derived from Argentina's Prometea)",,"Corte Constitucional de Colombia with the Universidad de Buenos Aires (IALAB, Juan Corvalán) and the Universidad del Rosario",Colombia,Corte Constitucional de Colombia,Colombia,"Bogotá, Colombia",Colombia,Justice administration: constitutional review case selection,Constitutional court case management,Public sector,Text and language,Public servants,16,Upper middle income,operational_improvement,"Faster data management, decision-making",,,,,,,,personal; organizational; societal,8,0,a person reading tutela files by hand to find the chamber's selection criterion,una persona tarda al menos un día para leer 30 expedientes,el sistema puede procesar las sentencias en medio digital en menos de 2 minutos,assumed,persisted,2024-08-02,not shown,https://www.dejusticia.org/wp-content/uploads/2021/04/CPC_informe_Colombia.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Corte Constitucional de Colombia, Sala Segunda de Revisión, Sentencia T-323 de 2024, 2 August 2024. The court that operates PretorIA wrote this ruling on the use of ChatGPT by a lower-court judge. In it, the court describes PretorIA as a tool that it applies and states that the tool neither selects nor prioritizes cases."", ""url"": ""https://www.corteconstitucional.gov.co/relatoria/2024/T-323-24.htm""}]","{""kind"": ""Official documents""}" +AIBD-0334,2020-07-27,source publication,day,intended,"Colombia's Constitutional Court presents PretorIA (July 2020), an AI system that classifies tutela rulings for selection. It was in user testing in April 2021. It was still in use in 2024, with no published measure of an outcome","Colombia's Constitutional Court presented PretorIA on 27 July 2020 to classify tutela rulings for selection. It was still in use in 2024, with no published measure of outcomes.",Colombia's Constitutional Court adopts PretorIA to sort tutela rulings,,,reported,Independently reported,,very low,program,false,"PretorIA (Corte Constitucional tutela classification and search system, derived from Argentina's Prometea)",,"Corte Constitucional de Colombia with the Universidad de Buenos Aires (IALAB, Juan Corvalán) and the Universidad del Rosario",Colombia,Corte Constitucional de Colombia,Colombia,"Bogotá, Colombia",Colombia,Justice administration: constitutional review case selection,Constitutional court case management,Public sector,Text and language,Public servants,16,Upper middle income,operational_improvement,"Faster data management, decision-making",,,,,,,,personal; organizational; societal,8,0,a person reading tutela files by hand to find the chamber's selection criterion,una persona tarda al menos un día para leer 30 expedientes,el sistema puede procesar las sentencias en medio digital en menos de 2 minutos,assumed,persisted,2024-08-02,not shown,https://www.dejusticia.org/wp-content/uploads/2021/04/CPC_informe_Colombia.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Court document"", ""value"": ""Corte Constitucional de Colombia, Sala Segunda de Revisión, Sentencia T-323 de 2024, 2 August 2024. The court that operates PretorIA wrote this ruling on the use of ChatGPT by a lower-court judge. In it, the court describes PretorIA as a tool that it applies and states that the tool neither selects nor prioritizes cases."", ""url"": ""https://www.corteconstitucional.gov.co/relatoria/2024/T-323-24.htm""}]","{""kind"": ""Official documents""}" AIBD-0748,2019-11-01,source publication,month,tested,Assessors for AHRQ find that Abstrackr's predictions caught 33–90% of records that a single reviewer wrongly excluded in screening simulations,"In simulations on three completed reviews, Abstrackr caught 33–90% of records that a single reviewer wrongly excluded. It cut screening workload by a median 40%.",*Abstrackr* catches studies that a lone reviewer wrongly excluded,33–90% of records that a single reviewer wrongly excluded,,reported,Independently reported,tested against a standard,very low,aggregate result,false,Abstrackr (machine learning citation screening),,"Brown University, Center for Evidence Synthesis in Health",United States,AHRQ's University of Alberta Evidence-based Practice Center (retrospective simulations on three reviews),Canada,"University of Alberta, Canada",Canada,Evidence synthesis / systematic reviews,,Work and industry,Text and language,Researchers,3,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,3,0,"A single reviewer's screening without the tool, checked against dual independent screening",,33–90% of records erroneously excluded by a single reviewer identified,measured,persisted,2020-06-03,shown,https://europepmc.org/abstract/MED/31790164,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Gates A, et al. Performance and Usability of Machine Learning for Screening in Systematic Reviews: A Comparative Evaluation of Three Tools. AHRQ Methods for Effective Health Care, Report No. 19(20)-EHC027-EF. Agency for Healthcare Research and Quality, November 2019. NCBI Bookshelf NBK550175, PMID 31790164 (structured abstract). The authors are the University of Alberta team that ran Abstrackr in these simulations."", ""url"": ""https://europepmc.org/abstract/MED/31790164""}]","{""kind"": ""Official documents""}" AIBD-0030,2018-09-01,event,day,tested,"Machine-learning model detects more cardiac arrests than dispatchers on emergency calls in Copenhagen, but in a randomized trial its alerts bring no significant gain in recognition by dispatchers","On emergency calls in Copenhagen, Corti's model detected more cardiac arrests than dispatchers did. In a trial, its alerts did not significantly improve how often dispatchers recognized arrests.",Software detects more cardiac arrests than Copenhagen dispatchers do,higher sensitivity than dispatchers (84.1% versus 72.5%),,reported,Independently reported,tested against a standard,very low,aggregate result,false,"Corti machine-learning call-analysis framework for out-of-hospital cardiac arrest recognition (the peer-reviewed papers describe it only as a machine learning framework, and the EENA report names Corti)",,"Corti (Copenhagen, Denmark)",Denmark,"Copenhagen Emergency Medical Services, Emergency Medical Dispatch Center Copenhagen",Denmark,"Copenhagen, Denmark",Denmark,Emergency medical dispatch / public safety,112 emergency call handling,Health,Speech and sound; Prediction from data,Health workers; Patients,3,High income,physical_health,Medical diagnosis,,,,,,,,personal; organizational,9,2,"Dispatchers without alerts in the trial's control group, on confirmed cardiac arrests",77.5% sensitivity for dispatchers without alerts,85.0% sensitivity for the machine learning alerts alone,measured,withdrawn,2025-06-10,shown,https://pubmed.ncbi.nlm.nih.gov/36634755/,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Follow-up"", ""value"": ""Ingeniøren reported that the emergency dispatch centres had shelved the model and that they did not recognize Corti's account of its effect."", ""url"": ""https://ing.dk/artikel/hjertestop-i-112-opkald-skulle-opdages-med-ai-derfor-blev-det-droppet""}]","{""kind"": ""Withdrawals""}" AIBD-0314,2018-04-10,event,day,mixed,"In a trial, Delhi Police facial recognition matches 2,930 children in care homes, but the Delhi High Court records that it has not helped solve any missing-child case","The Wire identifies Delhi Police's facial recognition as Innefu Labs' AI Vision. It produced 2,930 matches among children in care. The High Court later recorded that it solved no case.","Delhi's face recognition matches 2,930 children without solving a case","produced 2,930 matches among children in care",,reported,Independently reported,no comparison,very low,aggregate result,false,Delhi Police Facial Recognition System (reported by The Wire as Innefu Labs' AI Vision),AI Vision,"Innefu Labs Pvt Ltd (manufacturer, per The Wire citing the BPR&D Compendium of Equipment); supplied by Pelorus Technologies Pvt Ltd",India,"Delhi Police (Crime Branch), using Ministry of Women and Child Development Track Child portal data",India,"Delhi, India",India,Policing / missing persons and child protection,Missing children in child-care institutions,Safety and rescue,Images and video,Children; Public servants,16,Lower middle income,physical_health,,,,,,2,,,personal; organizational; societal,10,2,,,,,reversed,2019-01-22,shown,https://web.archive.org/web/2026id_/https://pulitzercenter.org/stories/looking-back-camera-how-controversial-ai-firms-shaped-delhis-predictive-policing,,,,[],Querist,The benefit was later reversed.,"[{""label"": ""Court document"", ""value"": ""Delhi High Court (2019). Sadhan Haldar v The State NCT of Delhi & Ors, W.P.(CRL) 1560/2017, order of 22 January 2019 (Hima Kohli and Manoj Kumar Ohri JJ). Indian Kanoon."", ""url"": ""https://indiankanoon.org/doc/181795735/""}, {""label"": ""Court document"", ""value"": ""High Court of Delhi (2019). Sadhan Haldar v. The State NCT of Delhi & Ors, W.P.(CRL) 1560/2017, order dated 26 April 2019 (Hima Kohli and Manoj Kumar Ohri JJ.)."", ""url"": ""https://delhihighcourt.nic.in/app/downloadOrderbByDate/W.P.%28CRL%29/1560/2017/26-04-2019""}, {""label"": ""Audit report"", ""value"": ""Comptroller and Auditor General of India (Principal Accountant General (Audit), Delhi) (2023). Performance Audit Report of the Comptroller and Auditor General of India on Children in Need of Care and Protection for the year ended 31 March 2021, Government of National Capital Territory of Delhi, Report No. 1 of the year 2023, Chapter 3, paragraph 3.2 (Delay in taking action by CWCs to use Facial Recognition Software), report pages 13-14."", ""url"": ""https://cag.gov.in/webroot/uploads/download_audit_report/2023/Report-No.-1-of-2023_CNCP_English-_09-02-2023-069c26599054797.71061814.pdf""}, {""label"": ""Follow-up"", ""value"": ""In August 2018 police counsel had reported a match rate of two percent. On 22 January 2019 the Delhi High Court recorded in Sadhan Haldar v. NCT of Delhi that the facial recognition software had not helped crack any case of a missing child."", ""url"": ""https://indiankanoon.org/doc/181795735/""}]","{""kind"": ""Challenges""}" AIBD-0931,2017-01-01,event,year,tested,"Researchers in the NHS screening program find that EyeArt could halve the grading of 30,405 episodes of English eye screening without missing moderate-to-severe or proliferative disease","In 30,405 episodes of English eye screening, EyeArt flagged 95.7% of referable retinopathy. It flagged all disease from moderate to proliferative. Its grades would have cleared about half of episodes.",*EyeArt* could halve retinal grading and miss no high-risk retinopathy,"15,091 of 30,405, would still have needed human grading",,reported,Independently reported,tested against a standard,very low,aggregate result,false,EyeArt,v2.1.0,"Eyenuk, Inc.",United States,"Three NHS diabetic eye screening programs in England (North East London, South East London and Gloucestershire), in a research evaluation run in parallel with usual care",United Kingdom,"North East London, South East London and Gloucestershire, England, United Kingdom",United Kingdom,Healthcare (diabetic eye screening),,Health,Images and video,Patients,3,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,3,0,"The current pathway, in which human graders grade every screening episode",All 30 405 episodes graded by humans,15 091/30 405 episodes needing further human grading after EyeArt,measured,unknown,,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC8077216/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""UK National Screening Committee (UK NSC) (2026). UK NSC consults on use of artificial intelligence in the diabetic eye screening programme (UK NSC blog, on its 2025 evidence summary), Paragraph on the 2024 resubmission; 'The evidence' section."", ""url"": ""https://nationalscreening.blog.gov.uk/2026/05/11/uk-nsc-consults-on-use-of-artificial-intelligence-in-the-diabetic-eye-screening-programme/""}]","{""kind"": ""Official documents""}" AIBD-2157,2016-01-01,event,year,reported,LAURA,"LAURA reads charts in 13 Brazilian hospitals and alerts teams to sepsis. Its medical director reported 25% fewer deaths in a before-and-after study, and one hospital a fall from 4.33% to 1.64%.",Brazilian hospitals report fewer deaths after *LAURA*'s sepsis alerts,25% fewer deaths,,claimed,The deployer's account,before and after,very low,aggregate result,false,LAURA (alerts of sepsis and clinical deterioration from patient charts),,Laura (Brazil),Brazil,"Thirteen hospitals in three Brazilian states, by the count of October 2019",Brazil,"Paraná, Minas Gerais and Rio Grande do Sul, Brazil",Brazil,Health care (early warning of sepsis and clinical deterioration),,Health,Prediction from data,Patients; Health workers,3,Upper middle income,physical_health,,5,4,5,,3,,,personal,4,1,The six months before the hospitals began to use the platform,,mortality 25% lower and stays 10% shorter,measured,unknown,,shown,https://fairlac.iadb.org/piloto/laura,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Eticas Research and Consulting. Robot Laura: Auditoría Algorítmica. Estudio sobre el sistema Laura de predicción de riesgo de deterioro clínico. Inter-American Development Bank, December 2021. The audit examined acceptability, usability, data protection and fairness by sex and age. It did not measure deaths. Archived copy of 20 July 2023. The quotation is in Spanish."", ""url"": ""https://web.archive.org/web/20230720134033id_/https://publications.iadb.org/publications/spanish/document/Robot-laura-auditoria-algoritmica.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0064,2025-08-01,event,month,reported,"After a scan supported by Brainomix 360 Stroke identifies her major stroke, an 83-year-old woman leaves the Royal Berkshire Hospital for thrombectomy in Oxford within 25 minutes",A scan supported by Brainomix 360 Stroke identified a woman's major stroke on arrival. She was transferred in 25 minutes for successful thrombectomy. She recovered well.,"After a scan aided by Brainomix finds a stroke, she leaves in 25 minutes",transferred in 25 minutes for successful thrombectomy,,claimed,The deployer's account,,very low,benefit created,false,Brainomix 360 Stroke,,Brainomix,United Kingdom,NHS England / Royal Berkshire Hospital,United Kingdom,"Reading, United Kingdom",United Kingdom,Healthcare - acute stroke care,Hospital stroke imaging decision support,Health,Images and video,Patients; Older people,3,High income,physical_health,Medical diagnosis,4,1,,,4,1,,personal; organizational,5,1,"thrombectomy rate at the same 26 hospitals before the software was implemented, with the 81 non-implementing hospitals as the background trend",2·3% (376 of 15 969 patients),"4·6% (751 of 15 428 patients), a relative increase of 100%",measured,scaled,2025-12-02,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf23/K231837.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence, Artificial intelligence (AI)-derived software to help clinical decision making in stroke, HealthTech guidance HTG708 (formerly DG57), 23 January 2024. NICE has no stake in the software. Its committee allowed e-Stroke in the NHS while more evidence is generated and found that the effect of the software on access to treatment is unclear."", ""url"": ""https://www.nice.org.uk/guidance/htg708/chapter/1-Recommendations""}]","{""kind"": ""Official documents""}" +AIBD-0064,2025-08-01,event,month,reported,"After a scan supported by Brainomix 360 Stroke identifies her major stroke, an 83-year-old woman leaves the Royal Berkshire Hospital for thrombectomy in Oxford within 25 minutes",A scan supported by Brainomix 360 Stroke identified a woman's major stroke on arrival. She was transferred in 25 minutes for successful thrombectomy. She recovered well.,"After a scan aided by Brainomix finds a stroke, she leaves in 25 minutes",transferred in 25 minutes for successful thrombectomy,,claimed,The deployer's account,no comparison,very low,benefit created,false,Brainomix 360 Stroke,,Brainomix,United Kingdom,NHS England / Royal Berkshire Hospital,United Kingdom,"Reading, United Kingdom",United Kingdom,Healthcare - acute stroke care,Hospital stroke imaging decision support,Health,Images and video,Patients; Older people,3,High income,physical_health,Medical diagnosis,4,1,,,4,1,,personal; organizational,5,1,"thrombectomy rate at the same 26 hospitals before the software was implemented, with the 81 non-implementing hospitals as the background trend",2·3% (376 of 15 969 patients),"4·6% (751 of 15 428 patients), a relative increase of 100%",measured,scaled,2025-12-02,shown,https://www.accessdata.fda.gov/cdrh_docs/pdf23/K231837.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""National Institute for Health and Care Excellence, Artificial intelligence (AI)-derived software to help clinical decision making in stroke, HealthTech guidance HTG708 (formerly DG57), 23 January 2024. NICE has no stake in the software. Its committee allowed e-Stroke in the NHS while more evidence is generated and found that the effect of the software on access to treatment is unclear."", ""url"": ""https://www.nice.org.uk/guidance/htg708/chapter/1-Recommendations""}]","{""kind"": ""Official documents""}" AIBD-0187,2025-06-30,source publication,day,reported,Federal analysts detect anomalous Medicare billing. $4.41 billion in claims is held in escrow,Justice Department analysts found anomalous Medicare billing through proactive data analytics. $4.41 billion was held in escrow. The software is unnamed. The department has not said that it used AI.,Medicare withholds $4.41 billion in suspected fraudulent payments,held $4.41 billion in escrow from about $4.45 billion scheduled for payment,"The Department of Justice reported on 30 June 2025 that its analysts and their collaborators had detected anomalous Medicare billing in Operation Gold Rush. According to the department, health agencies stopped all but about $41 million of about $4.45 billion scheduled for payment. The agencies were the Office of Inspector General of the Department of Health and Human Services and the Centers for Medicare & Medicaid Services. The $4.41 billion that was stopped went into escrow. The Justice Department said that the health agencies intended to seek its return to the Medicare trust fund. In its later announcements, the department does not say what finally became of that money.[1] Medicare supplemental insurers still paid about $900 million. The Justice Department credits the detection to proactive data analytics. In the same release, it announced plans for an AI-enabled Data Fusion Center. According to the American Hospital Association, accountable care organizations had also reported the suspect billing for intermittent urinary catheters. In June 2024 the Centers for Medicare & Medicaid Services called the 2023 volume of that billing highly suspect.[2] No source shows that AI flagged these claims. None says what share of the hold came from the analytics.[3]",claimed,The deployer's account,no comparison,very low,harm prevented,false,DOJ Health Care Fraud Unit proactive billing analytics,,"U.S. Department of Justice, Criminal Division Fraud Section, Health Care Fraud Unit Data Analytics Team",United States,U.S. Department of Justice with HHS-OIG and CMS,United States,United States,United States,Health care fraud and public insurance,Medicare durable medical equipment claims,Finance,Prediction from data,Public servants,16,High income,privacy_protection,Fraud detection,4,1,1,,3,1,,societal; organizational; personal,6,0,the amount Medicare was scheduled to pay on the fraudulent claims had the billing gone undetected,the approximately $4.45 billion that was scheduled to be paid by Medicare,all but approximately $41 million,measured,litigated,2026-08-13,not shown,https://www.justice.gov/opa/pr/national-health-care-fraud-takedown-results-324-defendants-charged-connection-over-146,$4.41 billion,"fraudulent Medicare payments held in escrow rather than paid out, to be returned to the Medicare trust fund",False,[],Querist,The use of the system was taken to court.,"[{""label"": ""Court document"", ""value"": ""United States v. [name withheld] et al., No. 25-cr-203, indictment, 18 June 2025, 1–10, ¶¶ 1–20, 32–34 (E.D.N.Y.)."", ""url"": ""https://www.justice.gov/criminal/media/1405326/dl?inline=""}, {""label"": ""Follow-up"", ""value"": ""DOJ reported 35 people charged and 16 convicted by 14 August 2026, after a guilty plea to money laundering the previous day. Neither what AI contributed nor what finally happened to the $4.41 billion in escrow is known."", ""url"": ""https://www.justice.gov/opa/pr/ohio-man-pleads-guilty-laundering-health-care-fraud-proceeds-transnational-criminal""}]","{""kind"": ""Legal proceedings""}" AIBD-0139,2024-04-02,event,day,reported,Tesla Full Self-Driving v12 may have driven its owner 13 miles to a Veterans Affairs emergency room during an insulin-pump failure at 2 a.m. in North Carolina,"By his own uncorroborated account on X, Tesla Full Self-Driving (Supervised) v12 drove its owner 13 miles to a VA emergency room at 2 a.m. on 2 April 2024.",Tesla software may have driven a sick owner 13 miles to a hospital,drove its owner 13 miles to a VA emergency room,,claimed,The beneficiary's account,no comparison,very low,protective response,false,Tesla Full Self-Driving (Supervised) v12 on a Tesla Model Y,"FSD v12 (Supervised), unlocked on the car on 1 April 2024 per the owner and Tesla Oracle; the exact build is not stated (The Driven describes the then-current release as 12.3.3)",Tesla,United States,,,"Asheville, United States",United States,Road transport / access to emergency care,Consumer driver-assistance software on a privately owned vehicle,Safety and rescue,Machines that act,Drivers and travelers; Patients,3,High income,physical_health,Medical access,4,1,1,,5,1,,personal,7,2,driving himself while impaired or calling emergency services,,"reached the VA emergency room 13 miles away, on his account without intervening",assumed,challenged,2024-05-14,shown,https://en.wikipedia.org/wiki/Tesla_Autopilot,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Regulatory document"", ""value"": ""National Highway Traffic Safety Administration, Office of Defects Investigation, Information Request PE24031-01 to Tesla, Inc., 5 November 2024, Enclosure 3 (email to Tesla of 14 May 2024). NHTSA regulates vehicle safety and has no stake in the owner's story. Its staff cite Tesla's repost of it as an example of a message that may lead drivers to treat FSD as a chauffeur. They do not assess whether the drive happened."", ""url"": ""https://static.nhtsa.gov/odi/inv/2024/INIM-PE24031-27346.pdf""}, {""label"": ""Follow-up"", ""value"": ""NHTSA raised the owner's story with Tesla on 15 April 2024 and in writing on 14 May 2024. It said that Tesla's repost conflicted with Tesla's own message that drivers must stay in control. NHTSA published the email with its information request of 5 November 2024. NHTSA questioned the promotion of the story and did not rule on whether the drive took place. No hospital or press confirmation of the owner's account has appeared since."", ""url"": ""https://static.nhtsa.gov/odi/inv/2024/INIM-PE24031-27346.pdf""}]","{""kind"": ""Challenges""}" -AIBD-1538,2020-12-31,event,year,reported,Sentry airstrike warnings for civilians in northwest Syria,Hala Systems estimates that its Sentry warnings cut deaths from airstrikes in northwest Syria by 20% in 2019 and 2020. Evaluators for a funder report the estimate and did not validate it.,*Sentry* warns Syrians of airstrikes minutes before they land,20%,,claimed,The evaluator's account,,very low,aggregate result,false,Sentry early warning system,,Hala Systems,United States,Hala Systems with the Syrian Civil Defence (White Helmets),Syria,Northwest Syria,Syria,Humanitarian protection (airstrike early warning),,Humanitarian,Prediction from data; Text and language,The public,16,Low income,physical_health,Hazard reduction,4,5,,4,3,,,personal; organizational; societal,2,0,"Areas and periods without timely warnings, in Hala Systems' own analysis",,20% lower mortality in 2019 and 2020,measured,unknown,,shown,https://aiforsocialgood.github.io/neurips2019/accepted/track1/posters/73_aisg_neurips2019.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Hale S, Poulson C, Winslow E, Mahrousa MB, Alhayes M. Outcome case study: Hala Systems. Triple Line Consulting for Creating Hope in Conflict: A Humanitarian Grand Challenge, December 2023. The funders of Hala Systems' grant commissioned the study. The evaluators report the developer's casualty estimates and state that they did not validate its method."", ""url"": ""https://alnap.org/documents/22146/Hala-Systems-Outcome-Case-Study.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Hale S, Poulson C, Winslow E, Mahrousa MB, Alhayes M. Outcome case study: Hala Systems. Triple Line Consulting for Creating Hope in Conflict: A Humanitarian Grand Challenge, December 2023. The funders of Hala Systems' grant commissioned the study. The evaluators report the developer's casualty estimates and state that they did not validate its method."", ""url"": ""https://alnap.org/documents/22146/Hala-Systems-Outcome-Case-Study.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Hale S, Poulson C, Winslow E, Mahrousa MB, Alhayes M. Outcome case study: Hala Systems. Triple Line Consulting for Creating Hope in Conflict: A Humanitarian Grand Challenge, December 2023. The funders of Hala Systems' grant commissioned the study. The evaluators report the developer's casualty estimates and state that they did not validate its method."", ""url"": ""https://alnap.org/documents/22146/Hala-Systems-Outcome-Case-Study.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0269,2025-11-18,source information,day,reported,An 84-year-old woman with no heart symptoms has valve replacement for severe aortic stenosis that echocardiography shows after EchoNext flags her pre-operative ECG,"At Columbia, the EchoNext deep-learning model flagged an 84-year-old's ECG before surgery. Echocardiography then showed severe aortic stenosis. She had a valve replacement after her operation.",Doctors find valve disease after EchoNext flags a pre-surgery ECG,an ejection fraction of 40% and severe aortic stenosis,,claimed,The makers' claim,,very low,benefit created,false,EchoNext deep-learning ECG model integrated into the electronic health record,second-generation EchoNext model,"Columbia University (Poterucha, Elias and colleagues)",United States,Columbia University Irving Medical Center / NewYork-Presbyterian,United States,"New York, New York, USA",United States,Health; cardiology,Hospital patients with an ECG and no known structural heart disease,Science,Prediction from data,Patients; Older people,3,High income,physical_health,Medical diagnosis,3,1,5,,4,1,,personal; organizational; societal,6,1,"cardiologists reading the same ECGs without AI assistance (1,600 reviews of 150 ECGs)",the accuracy of the cardiologists was 64.0% (95% CI 61.6–66.4%) with sensitivity 61.1% (95% CI 57.3–64.8%) and specificity 66.1% (95% CI 63.0–69.1%),"an accuracy of 77.3%, sensitivity 72.6% and specificity of 80.7%",measured,scaled,2026-06-08,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC12328201/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration. 510(k) K260332, EchoNext, Pathwai, Inc.: decision letter of 8 June 2026 and the sponsor's 510(k) summary. The summary is written by the sponsor, whose chief executive is the developer Pierre Elias. It covers the validation that FDA reviewed and shows that FDA cleared the device as substantially equivalent to the Viz.ai HCM module."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf26/K260332.pdf""}]","{""kind"": ""Official documents""}" +AIBD-1538,2020-12-31,event,year,reported,Sentry airstrike warnings for civilians in northwest Syria,Hala Systems estimates that its Sentry warnings cut deaths from airstrikes in northwest Syria by 20% in 2019 and 2020. Evaluators for a funder report the estimate and did not validate it.,*Sentry* warns Syrians of airstrikes minutes before they land,20%,,claimed,The evaluator's account,controlled comparison,low,aggregate result,false,Sentry early warning system,,Hala Systems,United States,Hala Systems with the Syrian Civil Defence (White Helmets),Syria,Northwest Syria,Syria,Humanitarian protection (airstrike early warning),,Humanitarian,Prediction from data; Text and language,The public,16,Low income,physical_health,Hazard reduction,4,5,,4,3,,,personal; organizational; societal,2,0,"Areas and periods without timely warnings, in Hala Systems' own analysis",,20% lower mortality in 2019 and 2020,measured,unknown,,shown,https://aiforsocialgood.github.io/neurips2019/accepted/track1/posters/73_aisg_neurips2019.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Hale S, Poulson C, Winslow E, Mahrousa MB, Alhayes M. Outcome case study: Hala Systems. Triple Line Consulting for Creating Hope in Conflict: A Humanitarian Grand Challenge, December 2023. The funders of Hala Systems' grant commissioned the study. The evaluators report the developer's casualty estimates and state that they did not validate its method."", ""url"": ""https://alnap.org/documents/22146/Hala-Systems-Outcome-Case-Study.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Hale S, Poulson C, Winslow E, Mahrousa MB, Alhayes M. Outcome case study: Hala Systems. Triple Line Consulting for Creating Hope in Conflict: A Humanitarian Grand Challenge, December 2023. The funders of Hala Systems' grant commissioned the study. The evaluators report the developer's casualty estimates and state that they did not validate its method."", ""url"": ""https://alnap.org/documents/22146/Hala-Systems-Outcome-Case-Study.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Hale S, Poulson C, Winslow E, Mahrousa MB, Alhayes M. Outcome case study: Hala Systems. Triple Line Consulting for Creating Hope in Conflict: A Humanitarian Grand Challenge, December 2023. The funders of Hala Systems' grant commissioned the study. The evaluators report the developer's casualty estimates and state that they did not validate its method."", ""url"": ""https://alnap.org/documents/22146/Hala-Systems-Outcome-Case-Study.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0269,2025-11-18,source information,day,reported,An 84-year-old woman with no heart symptoms has valve replacement for severe aortic stenosis that echocardiography shows after EchoNext flags her pre-operative ECG,"At Columbia, the EchoNext deep-learning model flagged an 84-year-old's ECG before surgery. Echocardiography then showed severe aortic stenosis. She had a valve replacement after her operation.",Doctors find valve disease after EchoNext flags a pre-surgery ECG,an ejection fraction of 40% and severe aortic stenosis,,claimed,The makers' claim,no comparison,very low,benefit created,false,EchoNext deep-learning ECG model integrated into the electronic health record,second-generation EchoNext model,"Columbia University (Poterucha, Elias and colleagues)",United States,Columbia University Irving Medical Center / NewYork-Presbyterian,United States,"New York, New York, USA",United States,Health; cardiology,Hospital patients with an ECG and no known structural heart disease,Science,Prediction from data,Patients; Older people,3,High income,physical_health,Medical diagnosis,3,1,5,,4,1,,personal; organizational; societal,6,1,"cardiologists reading the same ECGs without AI assistance (1,600 reviews of 150 ECGs)",the accuracy of the cardiologists was 64.0% (95% CI 61.6–66.4%) with sensitivity 61.1% (95% CI 57.3–64.8%) and specificity 66.1% (95% CI 63.0–69.1%),"an accuracy of 77.3%, sensitivity 72.6% and specificity of 80.7%",measured,scaled,2026-06-08,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC12328201/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration. 510(k) K260332, EchoNext, Pathwai, Inc.: decision letter of 8 June 2026 and the sponsor's 510(k) summary. The summary is written by the sponsor, whose chief executive is the developer Pierre Elias. It covers the validation that FDA reviewed and shows that FDA cleared the device as substantially equivalent to the Viz.ai HCM module."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/pdf26/K260332.pdf""}]","{""kind"": ""Official documents""}" AIBD-0065,2024-10-16,source information,day,reported,"After Viz HCM flags hypertrophic cardiomyopathy in a 61-year-old man at Cleveland Clinic, mavacamten eases his symptoms",Viz HCM gave a 61-year-old man's electrocardiogram a probability above 0.95 for hypertrophic cardiomyopathy. Imaging confirmed the disease. Mavacamten eased his symptoms by week 8.,Viz HCM flags hypertrophic cardiomyopathy in a 61-year-old,by week 8 the resting gradient was 9 mm Hg,,claimed,The deployer's account,no comparison,very low,benefit created,false,Viz HCM,,Viz.ai,United States,Cleveland Clinic,United States,"Cleveland, United States",United States,Healthcare - cardiology,AI-ECG screening in a health system,Health,Prediction from data,Patients,3,High income,physical_health,Medical diagnosis,3,1,2,,4,1,,personal,4,1,the clinical reading of the same baseline ECG,interpreted clinically as normal sinus rhythm with first-degree atrioventricular block,a high probability score (>0.95) for HCM,measured,unknown,,shown,https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230003.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""U.S. Food and Drug Administration, De Novo Classification Request for Viz HCM (DEN230003), decision summary, granted 3 August 2023. FDA reviewers wrote this summary. They establish that the device flags signs of HCM on 12-lead ECGs with a low positive predictive value, and they say nothing about this patient."", ""url"": ""https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230003.pdf""}]","{""kind"": ""Official documents""}" AIBD-0080,2023-04-01,event,month,reported,Five adults without respiratory symptoms are diagnosed with early-stage lung cancer after qXR flags suspicious nodules on chest X-rays taken for non-respiratory reasons,"Qure.ai's qXR flagged lung nodules on chest X-rays taken for other reasons. Computed tomography and biopsy confirmed early-stage lung cancer in five adults, and four had surgery.",Five early cancers are found after qXR flags nodules on unrelated X-rays,"confirmed early-stage lung cancer in five adults, and four had surgery",,claimed,The deployer's account,no comparison,very low,benefit created,false,qXR,,Qure.ai,India,"Participating centers of the CREATE study, not named in the paper; the clinical authors are affiliated with Hacettepe University (Ankara) and with Aarthi Scans and Sri Ramachandra Medical College and Hospital (Chennai)",,,,Healthcare - thoracic radiology / lung cancer,AI on routine chest X-rays in a multicentre study,Health,Images and video,Patients,3,,physical_health,Medical diagnosis,3,2,,,5,1,,personal,3,2,routine chest X-ray reading without the AI,would likely have been missed or delayed,early-stage lung cancer in all five cases,assumed,unknown,,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC13058629/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Austrian Institute for Health Technology Assessment (AIHTA). AI-supported Chest X-Ray Analysis for Lung Cancer Detection: Systematic Review of clinical outcomes and organisational implications. AIHTA Project Report No. 171b. Vienna, February 2026."", ""url"": ""https://eprints.aihta.at/1598/1/HTA-Projektbericht_Nr.171b.pdf""}, {""label"": ""Assessment report"", ""value"": ""Austrian Institute for Health Technology Assessment (AIHTA). AI-supported Chest X-Ray Analysis for Lung Cancer Detection: Systematic Review of clinical outcomes and organisational implications. AIHTA Project Report No. 171b. Vienna, February 2026."", ""url"": ""https://eprints.aihta.at/1598/1/HTA-Projektbericht_Nr.171b.pdf""}]","{""kind"": ""Official documents""}" AIBD-0694,2023-01-01,event,year,reported,"Evaluators for UNDP credit AI-assisted approaches with canceling over 21,500 km² of Ukrainian farmland from the hazard database in 2023","Evaluators of UNDP's mine action project found that AI-assisted approaches let Ukraine cancel over 21,500 km² of farmland from its hazard database in 2023.","AI helps Ukraine remove 21,500 km² of farmland from its hazard records","over 21,500 km² of farmland",,claimed,The evaluator's account,no comparison,very low,aggregate result,false,"AI-assisted, satellite-imagery and machine-learning approach to cancelling and prioritizing suspected hazardous areas in Ukraine's mine action database (unnamed tool)",,Not named in the evaluation (UNDP research and innovation work under the Mine Action Project),,"Ukraine's mine action authorities (Secretariat of the National Mine Action Authority, Ministry of Economy, Environment and Agriculture and State Emergency Service) with UNDP",Ukraine,Ukraine (national),Ukraine,Humanitarian mine action,,Humanitarian,Images and video,Farmers,2,Upper middle income,physical_health; operational_improvement,Hazard reduction,3,,,4,3,1,,societal; organizational,2,0,,,,,unknown,,shown,https://www.unevaluation.org/get-download-file/18787,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Shubitidze, G. Midterm Evaluation of the project 'United Nations Development Programme (UNDP) Mine Action Project (MAP) in Ukraine'. Evaluation report commissioned by UNDP Ukraine, 30 December 2025. Published in the UN Evaluation Group repository."", ""url"": ""https://www.unevaluation.org/get-download-file/18787""}]","{""kind"": ""Official documents""}" @@ -450,7 +450,7 @@ AIBD-0085,2023-05-17,event,day,reported,STF's VitorIA identifies at least 247 si AIBD-0695,2023-01-01,event,year,mixed,"In a randomized evaluation for UNICEF, dropout fell by almost half in Mexican schools that used the AI-personalized MatematIA platform","In a randomized evaluation for UNICEF in 14 schools in the State of Mexico, dropout fell by almost 50 percent where students used MatematIA. Mathematics grades did not improve.",Dropout nearly halves in Mexican schools using *MatematIA*,dropout fell by almost 50 percent,,claimed,The evaluator's account,randomized trial,moderate,aggregate result,false,"MatematIA (AI-personalized, gamified remedial mathematics platform)",,UNICEF with Pixframe Studios,Mexico,UNICEF Mexico with the Secretaría de Educación Pública and CECyTEM (Colegio de Estudios Científicos y Tecnológicos del Estado de México),Mexico,"State of Mexico, Mexico (14 CECyTEM upper-secondary schools)",Mexico,Education,,Education,Prediction from data,Students,4,Upper middle income,opportunity_access; incentivization,Educational access,2,3,3,2,2,,,personal,1,0,Students in the 7 randomized control schools without the platform,Average dropout rate of 1.58 percent across the sample,0.92 percentage points lower in the treatment schools,measured,unknown,,not shown,https://www.unevaluation.org/get-download-file/3824,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""EASE-ECONOMETRÍA for UNICEF Mexico. Uso de la plataforma MatematIA en CECyTEMs: Informe final ajustado (evaluation of the impact, effectiveness, relevance and sustainability of the MatematIA digital education platform), 27 November 2023. Commissioned by UNICEF and published in UNICEF's and the UN Evaluation Group's evaluation repositories."", ""url"": ""https://www.unevaluation.org/get-download-file/3824""}]","{""kind"": ""Official documents""}" AIBD-1149,2023-01-01,event,year,reported,Thunder Bay Electronics' evaluator scored the broadcaster's ENCO enCaption4 live captions mostly at or above Canada's 98 standard in 2023-2024,"The broadcaster's certified evaluator scored its automatic ENCO enCaption4 live news captions from 97.65 to 98.61 across 48 samples in 2023-2024, mostly at or above the CRTC's 98.",*enCaption4* captions mostly meet Canada's live accuracy standard,from 97.65 to 98.61 across 48 samples,,claimed,The evaluator's account,no comparison,very low,aggregate result,false,ENCO enCaption4 automated live captioning,enCaption4,ENCO Systems,United States,"Thunder Bay Electronics Limited (Dougall Media), for CKPR-DT and CHFD-DT",Canada,"Thunder Bay, Ontario, Canada",Canada,Broadcasting / live closed captions for deaf and hard-of-hearing viewers,,Accessibility,Speech and sound,,10,High income,accessibility,,2,,4,4,3,,,personal,3,0,,,,,unknown,,shown,https://enco.com/blog/enco-to-unveil-automated-closed-captioning-breakthroughs-at-nab-show-new-york,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Thunder Bay Electronics Limited (Dougall Media), CKPR-DT and CHFD-DT. Annual closed captioning report for the 2023/2024 broadcast season, filed with the CRTC under Broadcasting Regulatory Policy CRTC 2019-308, 29 November 2024, with NER scores by a certified NER evaluator that the broadcaster engaged. Archived copy of https://crtc.gc.ca/eng/BCASTING/ann_rep/tbayelec_cc2024.pdf."", ""url"": ""https://web.archive.org/web/20250425053207id_/https://crtc.gc.ca/eng/BCASTING/ann_rep/tbayelec_cc2024.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Thunder Bay Electronics Limited (Dougall Media), CKPR-DT and CHFD-DT. Annual closed captioning report for the 2023/2024 broadcast season, filed with the CRTC under Broadcasting Regulatory Policy CRTC 2019-308, 29 November 2024, with NER scores by a certified NER evaluator that the broadcaster engaged. Archived copy of https://crtc.gc.ca/eng/BCASTING/ann_rep/tbayelec_cc2024.pdf."", ""url"": ""https://web.archive.org/web/20250425053207id_/https://crtc.gc.ca/eng/BCASTING/ann_rep/tbayelec_cc2024.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Thunder Bay Electronics Limited (Dougall Media), CKPR-DT and CHFD-DT. Annual closed captioning report for the 2023/2024 broadcast season, filed with the CRTC under Broadcasting Regulatory Policy CRTC 2019-308, 29 November 2024, with NER scores by a certified NER evaluator that the broadcaster engaged. Archived copy of https://crtc.gc.ca/eng/BCASTING/ann_rep/tbayelec_cc2024.pdf."", ""url"": ""https://web.archive.org/web/20250425053207id_/https://crtc.gc.ca/eng/BCASTING/ann_rep/tbayelec_cc2024.pdf""}, {""label"": ""Regulatory document"", ""value"": ""Canadian Radio-television and Telecommunications Commission. Broadcasting Regulatory Policy CRTC 2019-308: English-language closed captioning mandatory quality standards relating to the accuracy rate for live television programming. Ottawa, 30 August 2019. Archived copy of https://crtc.gc.ca/eng/archive/2019/2019-308.htm. Appendix 1, Quality Standard 3."", ""url"": ""https://web.archive.org/web/20260919084435id_/https://crtc.gc.ca/eng/archive/2019/2019-308.htm""}, {""label"": ""Regulatory document"", ""value"": ""Canadian Radio-television and Telecommunications Commission. Broadcasting Regulatory Policy CRTC 2019-308: English-language closed captioning mandatory quality standards relating to the accuracy rate for live television programming. Ottawa, 30 August 2019. Archived copy of https://crtc.gc.ca/eng/archive/2019/2019-308.htm. Section on requested flexibility for independent broadcasters, relaying the joint intervention of the English-Language Local Independent Television Stations Group (E-ILTS)."", ""url"": ""https://web.archive.org/web/20260919084435id_/https://crtc.gc.ca/eng/archive/2019/2019-308.htm""}, {""label"": ""Regulatory document"", ""value"": ""Canadian Radio-television and Telecommunications Commission. Broadcasting Regulatory Policy CRTC 2019-308: English-language closed captioning mandatory quality standards relating to the accuracy rate for live television programming. Ottawa, 30 August 2019. Archived copy of https://crtc.gc.ca/eng/archive/2019/2019-308.htm. Footnote 6, the members of E-ILTS."", ""url"": ""https://web.archive.org/web/20260919084435id_/https://crtc.gc.ca/eng/archive/2019/2019-308.htm""}]","{""kind"": ""Official documents""}" AIBD-1218,2022-12-15,event,day,reported,A Duke Law evaluation shows that ShotSpotter cut Durham police's median response time to gunfire by 1.2 minutes relative to the rest of the city,"In Durham's 12-month pilot, evaluators found that median police response time fell 1.2 minutes more in the ShotSpotter area than elsewhere. They could not show an effect on gun violence.",*ShotSpotter* sped Durham's police response to gunfire by 1.2 minutes,a decline of 1.2 minutes in the target area compared with the rest of the city,,claimed,The evaluator's account,controlled comparison,low,aggregate result,false,ShotSpotter gunfire-detection service,,"SoundThinking, Inc.",United States,Durham Police Department (City of Durham),United States,"Durham, North Carolina, United States",United States,Policing (gunfire detection),,Safety and rescue,Speech and sound,Public servants,16,High income,physical_health; operational_improvement,Crime reduction or prevention,2,1,4,4,3,5,0.440,organizational; societal,2,0,"The rest of Durham, before and after the pilot",406 seconds (6.8 minutes) if the target area had followed the control pattern,"332 seconds (5.5 minutes), the observed 2023 median",measured,withdrawn,2024-03-05,shown,https://www.soundthinking.com/blog/soundthinkings-approach-to-fourth-of-july-safety/,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Evaluation report"", ""value"": ""Cook, Philip J. and Adam Soliman. Evaluation of Durham's ShotSpotter Installation: Results of a 12-Month Pilot Project. Wilson Center for Science and Justice, Duke Law, February 2024. The Durham Police Department arranged the evaluation."", ""url"": ""https://wcsj.law.duke.edu/wp-content/uploads/2024/02/ShotSpotter-Evaluation.pdf""}, {""label"": ""Follow-up"", ""value"": ""In March 2024, a week after the Duke evaluation, the Durham City Council voted 4–2 not to renew its ShotSpotter contract."", ""url"": ""https://abc11.com/post/durham-shotspotter-crime-gun-shots-fired/14491996/""}]","{""kind"": ""Withdrawals""}" -AIBD-0519,2022-02-01,event,month,mixed,"AI that interprets in sign language, for digital access",Baidu reports that its XiLing avatar gave round-the-clock sign-language coverage of the 2022 Winter Olympics. A Deaf professor's team found that the avatars lost information. It found distortions.,Baidu says *XiLing* signed round the clock at the Winter Olympics,,,claimed,The makers' claim,,very low,aggregate result,false,Baidu XiLing AI sign-language interpretation platform,,Baidu,,China Central Television (CCTV),,"2022 Winter Olympics, China",China,Accessibility,,Accessibility,Speech and sound,People with disabilities,10,Upper middle income,accessibility,,2,,2,2,2,,,personal; organizational,3,2,"The original spoken audio of the broadcasts, compared with a back-translation of the avatar signing",,Key information lost or distorted in the avatar signing,measured,challenged,2025-07-09,shown,https://www.itu.int/dms_pub/itu-t/opb/ai4g/T-AI4G-AI4GOOD-2025-5-PDF-E.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""A Deaf professor reported that a study of the Winter Olympics avatars by the professor's team found information loss and poor comprehension among Deaf viewers. The professor wrote that many Deaf Chinese and sign language interpreters had been skeptical of or resistant to the avatars."", ""url"": ""https://www.sixthtone.com/news/1017316""}]","{""kind"": ""Challenges""}" +AIBD-0519,2022-02-01,event,month,mixed,"AI that interprets in sign language, for digital access",Baidu reports that its XiLing avatar gave round-the-clock sign-language coverage of the 2022 Winter Olympics. A Deaf professor's team found that the avatars lost information. It found distortions.,Baidu says *XiLing* signed round the clock at the Winter Olympics,,,claimed,The makers' claim,no comparison,very low,aggregate result,false,Baidu XiLing AI sign-language interpretation platform,,Baidu,,China Central Television (CCTV),,"2022 Winter Olympics, China",China,Accessibility,,Accessibility,Speech and sound,People with disabilities,10,Upper middle income,accessibility,,2,,2,2,2,,,personal; organizational,3,2,"The original spoken audio of the broadcasts, compared with a back-translation of the avatar signing",,Key information lost or distorted in the avatar signing,measured,challenged,2025-07-09,shown,https://www.itu.int/dms_pub/itu-t/opb/ai4g/T-AI4G-AI4GOOD-2025-5-PDF-E.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""A Deaf professor reported that a study of the Winter Olympics avatars by the professor's team found information loss and poor comprehension among Deaf viewers. The professor wrote that many Deaf Chinese and sign language interpreters had been skeptical of or resistant to the avatars."", ""url"": ""https://www.sixthtone.com/news/1017316""}]","{""kind"": ""Challenges""}" AIBD-1135,2021-01-01,event,year,reported,EPRI reports that Pano AI's camera platform detected 85% of test fires within 10 minutes of visible smoke in a 2021 pilot with PGE and PG&E,"EPRI reports that in a 2021 pilot with utilities, Pano AI's platform detected 60% of test fires within 4 minutes of visible smoke. It detected 85% within 10 minutes.",*Pano AI* spots most test fires within minutes in a utility pilot,60% of test fires within 4 minutes of visible smoke,,claimed,The evaluator's account,tested against a standard,low,aggregate result,false,"Pano AI Rapid Detect platform (Pano Stations with AI-based smoke detection, and the Pano Intelligence Center)",final November production model,"Pano AI (San Francisco, California)",United States,"Portland General Electric (two Pano Stations) and Pacific Gas & Electric (13 stations), host utilities of the 2021 pilot",United States,"Service territories of Portland General Electric in Oregon and Pacific Gas & Electric in California, United States",United States,Wildfire detection / electric utilities,,Energy and infrastructure,Images and video,Public servants,13,High income,environmental_impact,Natural disaster prediction,2,2,3,,3,5,,organizational; societal,1,0,The platform's false positive rate earlier in the pilot,about 1.6% in late July,0.37% in early October,measured,scaled,2025-07-29,shown,https://restservice.epri.com/publicdownload/000000003002023034/0/Product,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""EPRI (Electric Power Research Institute). Incubatenergy Labs 2021 Pilot Project Report: 360-Degree, AI-Based Imaging for Wildfire Situational and Locational Awareness. Product 3002023034, March 2022. EPRI was the technical advisor to pilots hosted by Portland General Electric and Pacific Gas & Electric (evaluator's account)."", ""url"": ""https://restservice.epri.com/publicdownload/000000003002023034/0/Product""}, {""label"": ""Evaluation report"", ""value"": ""EPRI (Electric Power Research Institute). Incubatenergy Labs 2021 Pilot Project Report, March 2022, Results & Learnings (false positives)."", ""url"": ""https://restservice.epri.com/publicdownload/000000003002023034/0/Product""}]","{""kind"": ""Official documents""}" AIBD-0728,2020-01-01,event,year,mixed,"Stanford evaluators find that after Allegheny's LASSO screening model, with policy changes, screened-in children were less often removed from home","Stanford evaluators found that after Allegheny's LASSO screening model, with policy changes, the 180-day home removal risk after a screen-in fell 2.7 points. Acceptance for services fell 8 points.",Updated *Allegheny Family Screening Tool* precedes fewer home removals,fell by 2.7 percentage points,,claimed,The evaluator's account,controlled comparison,very low,aggregate result,false,"Allegheny Family Screening Tool, LASSO version (predictive risk model for child welfare hotline screening)",,Not named in the evaluation (built for the Allegheny County Department of Human Services),,"Allegheny County Department of Human Services, Office of Children, Youth and Families",United States,"Allegheny County, Pennsylvania, United States",United States,Child welfare,,Public sector,Prediction from data,Public servants; Children,16,High income,liberty_protection,,2,4,3,5,2,5,0.490,personal,2,0,"General protective service episodes before any tool, January 2015 to July 2016",,180-day risk of home removal after a screen-in 2.7 percentage points lower,measured,challenged,2023-06-12,shown,https://www.alleghenycountyanalytics.us/wp-content/uploads/2024/05/24-ACDHS-04-Predictive-Risk-Algorithms.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Evaluation report"", ""value"": ""Goldhaber-Fiebert JD, Prince L. Impact Evaluation of a Predictive Risk Modeling Tool for Allegheny County (Phase 2). Stanford University, 26 March 2023. Published by the Allegheny County Department of Human Services."", ""url"": ""https://analytics.alleghenycounty.us/wp-content/uploads/2024/05/23-ACDHS-19_FamilyScreeningToolUpdate_v2.pdf""}, {""label"": ""Follow-up"", ""value"": ""An audit by ACLU and HRDAG staff found that the tool, measured as the county uses it, had lower accuracy, more false positives and wider racial disparities than the county reported."", ""url"": ""https://www.aclu.org/the-devil-is-in-the-details-interrogating-values-embedded-in-the-allegheny-family-screening-tool""}]","{""kind"": ""Challenges""}" AIBD-1035,2020-01-01,event,year,reported,EPA reports that a machine-learning model for targeting hazardous-waste inspections detected violations more efficiently in a randomized field test,"EPA reports that a machine-learning model yielded a substantial increase in efficiency of detecting hazardous-waste violations in a randomized field test, with no extra operating cost.",*RCRA predictive targeting model* spots violations more efficiently,an 82 percent higher detection rate,,claimed,The makers' claim,randomized trial,high,aggregate result,false,"Machine-learning model that forecasts RCRA violations by large quantity generator (LQG) hazardous waste facilities, tested in the RCRA Data Analytics Evaluation",,"University of Chicago Energy & Environment Lab, with EPA staff",United States,"EPA regional offices, coordinated by EPA's Office of Enforcement and Compliance Assurance",United States,United States (EPA regions),United States,Environmental enforcement (hazardous waste),,Environment,Prediction from data,Public servants,12,High income,operational_improvement; environmental_impact,Enhanced productivity,2,1,4,,3,2,,organizational; societal,2,0,Status quo inspection targeting practices across EPA regions,,a substantial increase in efficiency of detecting hazardous waste violations,measured,scaled,,shown,https://www.epa.gov/system/files/documents/2023-06/fy20-epa-program-evaluations.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""U.S. Environmental Protection Agency. FY 2020 EPA Program Evaluations, Office of Enforcement and Compliance Assurance (OECA), Activity 1: RCRA Data Analytics Evaluation, p. 1. Evaluation run by EPA with the University of Chicago Energy & Environment Lab."", ""url"": ""https://www.epa.gov/system/files/documents/2023-06/fy20-epa-program-evaluations.pdf""}]","{""kind"": ""Official documents""}" @@ -470,23 +470,23 @@ AIBD-1196,2024-12-01,event,month,reported,The Government Digital Service's cross AIBD-1193,2024-10-01,event,month,reported,The Home Office's evaluation of its Asylum Case Summarisation tool shows that decision-makers reviewed transcripts 23 minutes faster per case,"In Home Office pilots in 2024, decision-makers using the Asylum Case Summarisation tool reviewed interview transcripts 23 minutes quicker per case than a comparison group, a 32% time saving.",*Asylum Case Summarisation* cuts transcript review by 23 minutes,23 minutes quicker per case,,claimed,The deployer's account,controlled comparison,very low,aggregate result,false,Asylum Case Summarisation (ACS) tool,,Home Office (the evaluation names no outside supplier),United Kingdom,Home Office asylum decision-makers in two Decision Making Units,United Kingdom,United Kingdom,United Kingdom,Asylum and immigration administration,,Public sector,Text and language,Public servants,10; 16,High income,operational_improvement,"Faster data management, decision-making",1,2,2,,3,,,organizational,2,0,A comparison group of decision-makers in the same Decision Making Units who worked without the tool,,23 minutes quicker per case (a 32% time saving),measured,challenged,2026-06-02,,,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Evaluation report"", ""value"": ""Home Office. Evaluation of AI trials in the asylum decision making process. Research and analysis, GOV.UK, 29 April 2025."", ""url"": ""https://www.gov.uk/government/publications/evaluation-of-ai-trials-in-the-asylum-decision-making-process/evaluation-of-ai-trials-in-the-asylum-decision-making-process""}, {""label"": ""Inquiry report"", ""value"": ""Independent Chief Inspector of Borders and Immigration. An inspection of asylum casework, June 2025 to December 2025. Presented to Parliament, June 2026. In the inspectorate's survey, 6 of 252 decision-makers had used the tool."", ""url"": ""https://assets.publishing.service.gov.uk/media/6a3cfd23eaee00074150f4a1/An_inspection_of_asylum_casework__June___December_2025_.pdf""}, {""label"": ""Follow-up"", ""value"": ""The Home Office rolled the tool out across asylum casework in 2026. On 2 June 2026 Open Rights Group, with a SOAS researcher, argued that it could be unlawful and that applicants are not told of its use."", ""url"": ""https://www.openrightsgroup.org/app/uploads/2026/06/AI-in-Asylum-Decisions-FINAL.pdf""}]","{""kind"": ""Challenges""}" AIBD-0393,2024-01-01,event,month,reported,AI Senior Care Contents for the well-being of seniors,ESTsoft offers AI Senior Care Content with conversational companions in Gimhae senior centers. It reports that older users described satisfaction with their overall well-being.,Older adults describe satisfaction with AI companions at senior centers,,,claimed,The makers' claim,no comparison,very low,aggregate result,false,ESTsoft AI Senior Care Content and conversational companion,,ESTsoft Inc.,,Gimhae City smart senior centers,,"Gimhae, South Korea; additional Japanese pilots",South Korea,Healthcare,,Health,Text and language,Older people,3,High income,emotional_health,Companionship,1,2,2,4,3,,,personal; organizational,3,1,,,,,unknown,,shown,https://www.itu.int/dms_pub/itu-t/opb/ai4g/T-AI4G-AI4GOOD-2025-5-PDF-E.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""The Korean Journal of Physical Education (한국체육학회지), Korean Alliance for Health, Physical Education, Recreation, and Dance (KAHPERD) (2024). AI 휴먼 시니어케어 헬스콘텐츠를 참여한 노인의 수용과정 탐색: 혁신확산이론을 중심으로 (Exploring the Acceptance of AI Human Senior Care Contents among Older Adults: Application of Innovation Diffusion Theory), Korean abstract, closing sentences; funding note on the title page. Vol. 63, no. 3, pp. 373-386, DOI 10.23949/kjpe.2024.5.63.3.28 (https://doi.org/10.23949/kjpe.2024.5.63.3.28). The DOI landing page (journal_kr.kahperd.or.kr) shows only the title; the url is the full-text JATS XML that the landing page loads from the publisher's KISS server."", ""url"": ""http://210.101.116.16/xml/9001/30803/03622614/03622614.xml""}]","{""kind"": ""Official documents""}" AIBD-0561,2024-01-01,event,year,reported,"South Cambridgeshire - a chatbot for the contact center, to improve resident satisfaction",South Cambridgeshire District Council reports that its EBI.AI chatbot cut calls. The chatbot improved routing. Councilors later said that some callers hung up without help.,South Cambridgeshire reports fewer calls after a chatbot,,,claimed,The deployer's account,before and after,very low,aggregate result,false,"EBI.AI chatbot, live chat, AI interactive voice response, and planned voicebot functions",,EBI.AI,,South Cambridgeshire District Council,,United Kingdom,United Kingdom,Council contact-center services,,Public sector,Text and language,Public servants; The public,16,High income,operational_improvement,Enhanced productivity,1,1,3,2,3,,,organizational; personal,2,0,"Call volumes in the same months of the previous year, before the chatbot","The previous year's call volumes, not stated","A noticeable reduction in call volumes, not quantified",measured,challenged,2026-06-09,shown,https://web.archive.org/web/2026id_/https://www.local.gov.uk/case-studies/south-cambridgeshire-contact-centre-chatbot-improve-resident-satisfaction,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""The system stayed in use on the web and by phone. At a scrutiny committee reported on 9 June 2026, an officer cited fewer than 10 complaints in a year and 87% positive webchat feedback. Councillors described residents who hung up without help and a dead-deer report that the system treated as fly-tipping."", ""url"": ""https://www.cambridge-news.co.uk/news/cambridge-news/four-day-week-council-says-34084986""}]","{""kind"": ""Challenges""}" -AIBD-0911,2022-01-01,event,year,mixed,"Pacific Northwest National Laboratory finds that alerts from SmartKit AI for demand response saved a GSA building $20,000 in 2022, while its alerts for load management failed","PNNL found that SmartKit AI's alerts for demand response cut a GSA building's utility charges by $20,000 for 2022. Staff could not act on its alerts for load management.",*SmartKit AI* alerts cut a Newark federal building's capacity charges,"$20,000 for 2022",,claimed,The evaluator's account,,very low,benefit created,false,SmartKit AI (operator-in-the-loop grid-interactive efficient building software),,Logical Buildings,United States,"U.S. General Services Administration, 20 Washington Place, Newark",United States,"Newark, New Jersey, United States",United States,Energy (building demand response),,Energy and infrastructure,Prediction from data,Public servants,7,High income,operational_improvement,Enhanced productivity,1,1,4,,2,3,,organizational,1,0,The building's average monthly maximum load from June to September 2022,449 kW [maximum],334 kW [average during ICAP events],measured,unknown,,shown,https://www.gsa.gov/system/files/GPG055-Operator%20in%20the%20Loop-508Compliant_Final%20Report.pdf,"$20,000",Savings in capacity and transmission (ICAP) charges for 2022 from demand response on the predicted days,True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Myer, M. Operator-in-the-Loop Grid-Interactive Efficient Building Evaluation. Pacific Northwest National Laboratory for the U.S. General Services Administration and the U.S. Department of Energy, GSA Proving Ground report GPG-055, April 2025."", ""url"": ""https://www.gsa.gov/system/files/GPG055-Operator%20in%20the%20Loop-508Compliant_Final%20Report.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Myer, M. Operator-in-the-Loop Grid-Interactive Efficient Building Evaluation. Pacific Northwest National Laboratory, GPG-055, April 2025. Executive summary."", ""url"": ""https://www.gsa.gov/system/files/GPG055-Operator%20in%20the%20Loop-508Compliant_Final%20Report.pdf""}]","{""kind"": ""Official documents""}" +AIBD-0911,2022-01-01,event,year,mixed,"Pacific Northwest National Laboratory finds that alerts from SmartKit AI for demand response saved a GSA building $20,000 in 2022, while its alerts for load management failed","PNNL found that SmartKit AI's alerts for demand response cut a GSA building's utility charges by $20,000 for 2022. Staff could not act on its alerts for load management.",*SmartKit AI* alerts cut a Newark federal building's capacity charges,"$20,000 for 2022",,claimed,The evaluator's account,no comparison,very low,benefit created,false,SmartKit AI (operator-in-the-loop grid-interactive efficient building software),,Logical Buildings,United States,"U.S. General Services Administration, 20 Washington Place, Newark",United States,"Newark, New Jersey, United States",United States,Energy (building demand response),,Energy and infrastructure,Prediction from data,Public servants,7,High income,operational_improvement,Enhanced productivity,1,1,4,,2,3,,organizational,1,0,The building's average monthly maximum load from June to September 2022,449 kW [maximum],334 kW [average during ICAP events],measured,unknown,,shown,https://www.gsa.gov/system/files/GPG055-Operator%20in%20the%20Loop-508Compliant_Final%20Report.pdf,"$20,000",Savings in capacity and transmission (ICAP) charges for 2022 from demand response on the predicted days,True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Myer, M. Operator-in-the-Loop Grid-Interactive Efficient Building Evaluation. Pacific Northwest National Laboratory for the U.S. General Services Administration and the U.S. Department of Energy, GSA Proving Ground report GPG-055, April 2025."", ""url"": ""https://www.gsa.gov/system/files/GPG055-Operator%20in%20the%20Loop-508Compliant_Final%20Report.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Myer, M. Operator-in-the-Loop Grid-Interactive Efficient Building Evaluation. Pacific Northwest National Laboratory, GPG-055, April 2025. Executive summary."", ""url"": ""https://www.gsa.gov/system/files/GPG055-Operator%20in%20the%20Loop-508Compliant_Final%20Report.pdf""}]","{""kind"": ""Official documents""}" AIBD-1217,2020-11-01,event,month,mixed,"Evaluators at Cleveland State University find that ShotSpotter alerts were located more accurately than 911 calls, letting Cleveland police respond faster","In a City-funded evaluation, Cleveland State criminologists found ShotSpotter alerts located more accurately than most 911 calls and answered faster. They say it alone does not cut crime.","*ShotSpotter* pinpoints Cleveland gunfire, getting officers there sooner",dispatched 81 seconds on average after an alert,,claimed,The evaluator's account,no comparison,very low,aggregate result,false,"ShotSpotter gunfire detection system (acoustic sensors, an automated algorithm and human review)",,"SoundThinking, Inc. (formerly ShotSpotter, Inc.)",United States,Cleveland Division of Police,United States,"Cleveland, Ohio, United States",United States,Policing (gunfire detection),,Safety and rescue,Speech and sound,Public servants,16,High income,physical_health; operational_improvement,Crime reduction or prevention,1,1,4,,2,5,,organizational; societal,2,0,911 calls reporting gunfire in Cleveland,,,assumed,persisted,2026-04-09,shown,https://www.soundthinking.com/law-enforcement/gunshot-detection-technology/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Kent, Stephanie L. and Rachel E. Lovell. Evaluation of the Cleveland Division of Police's Implementation of ShotSpotter. Criminology Research Center, Levin College of Public Affairs and Education, Cleveland State University, August 2025."", ""url"": ""https://signalcleveland.org/wp-content/uploads/2025/10/CSU-ShotSpotter-Final-Report-2025-09-05-FINAL-1.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Kent, Stephanie L. and Rachel E. Lovell. Evaluation of the Cleveland Division of Police's Implementation of ShotSpotter. Criminology Research Center, Levin College of Public Affairs and Education, Cleveland State University, August 2025."", ""url"": ""https://signalcleveland.org/wp-content/uploads/2025/10/CSU-ShotSpotter-Final-Report-2025-09-05-FINAL-1.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Kent, Stephanie L. and Rachel E. Lovell. Evaluation of the Cleveland Division of Police's Implementation of ShotSpotter. Criminology Research Center, Levin College of Public Affairs and Education, Cleveland State University, August 2025."", ""url"": ""https://signalcleveland.org/wp-content/uploads/2025/10/CSU-ShotSpotter-Final-Report-2025-09-05-FINAL-1.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Kent, Stephanie L. and Rachel E. Lovell. Evaluation of the Cleveland Division of Police's Implementation of ShotSpotter. Criminology Research Center, Levin College of Public Affairs and Education, Cleveland State University, August 2025."", ""url"": ""https://signalcleveland.org/wp-content/uploads/2025/10/CSU-ShotSpotter-Final-Report-2025-09-05-FINAL-1.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Kent, Stephanie L. and Rachel E. Lovell. Evaluation of the Cleveland Division of Police's Implementation of ShotSpotter. Criminology Research Center, Levin College of Public Affairs and Education, Cleveland State University, August 2025."", ""url"": ""https://signalcleveland.org/wp-content/uploads/2025/10/CSU-ShotSpotter-Final-Report-2025-09-05-FINAL-1.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0337,2020-01-01,event,year,reported,"Nearest-neighbor comparison chooses prompts in the Australian Taxation Office's myTax. They reach nearly 340,000 taxpayers at tax time 2020. The office estimates that the resulting adjustments had an impact of $37 million on revenue","At tax time 2020, nearest-neighbor checks in the Australian Taxation Office's myTax prompted nearly 340,000 users to review a claim. The office puts the impact on revenue at about $37 million.",Prompts in Australia's myTax lead taxpayers to adjust unusual claims,adjustments estimated to have a revenue impact of around $37 million,,claimed,The makers' claim,,very low,aggregate result,false,ATO myTax real-time analytics prompts using nearest-neighbor comparison models,,Australian Taxation Office (in-house data and analytics),Australia,Australian Taxation Office,Australia,"Australia (national, online myTax lodgment)",Australia,Public administration: tax compliance,National revenue agency,Finance,Prediction from data,The public,16,High income,operational_improvement; incentivization,Beneficial default actions,1,4,1,,4,2,,personal; organizational; societal,5,1,Taxpayers who did not receive a nudge message in the ATO's randomized control trial for the 2023-24 income year,,2.5 times more likely to review their return and 3 times more likely to adjust their data,measured,scaled,,shown,https://web.archive.org/web/20210714025448/https://www.ato.gov.au/About-ATO/Commitments-and-reporting/Information-and-privacy/How-we-use-data-and-analytics/,$37 million,revenue impact of the adjustments taxpayers made after myTax prompts at tax time 2020 (ATO's own estimate),True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Australian National Audit Office, Governance of Artificial Intelligence at the Australian Taxation Office, Auditor-General Report No. 26 of 2024–25, 24 February 2025. The audit office is independent of the ATO. Its auditors describe the real-time analytics component and find that the ATO did not document the weights behind it. They do not test its revenue effect."", ""url"": ""https://web.archive.org/web/20261014065948/https://www.anao.gov.au/work/performance-audit/governance-of-artificial-intelligence-the-australian-taxation-office""}]","{""kind"": ""Official documents""}" +AIBD-0337,2020-01-01,event,year,reported,"Nearest-neighbor comparison chooses prompts in the Australian Taxation Office's myTax. They reach nearly 340,000 taxpayers at tax time 2020. The office estimates that the resulting adjustments had an impact of $37 million on revenue","At tax time 2020, nearest-neighbor checks in the Australian Taxation Office's myTax prompted nearly 340,000 users to review a claim. The office puts the impact on revenue at about $37 million.",Prompts in Australia's myTax lead taxpayers to adjust unusual claims,adjustments estimated to have a revenue impact of around $37 million,,claimed,The makers' claim,no comparison,very low,aggregate result,false,ATO myTax real-time analytics prompts using nearest-neighbor comparison models,,Australian Taxation Office (in-house data and analytics),Australia,Australian Taxation Office,Australia,"Australia (national, online myTax lodgment)",Australia,Public administration: tax compliance,National revenue agency,Finance,Prediction from data,The public,16,High income,operational_improvement; incentivization,Beneficial default actions,1,4,1,,4,2,,personal; organizational; societal,5,1,Taxpayers who did not receive a nudge message in the ATO's randomized control trial for the 2023-24 income year,,2.5 times more likely to review their return and 3 times more likely to adjust their data,measured,scaled,,shown,https://web.archive.org/web/20210714025448/https://www.ato.gov.au/About-ATO/Commitments-and-reporting/Information-and-privacy/How-we-use-data-and-analytics/,$37 million,revenue impact of the adjustments taxpayers made after myTax prompts at tax time 2020 (ATO's own estimate),True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Audit report"", ""value"": ""Australian National Audit Office, Governance of Artificial Intelligence at the Australian Taxation Office, Auditor-General Report No. 26 of 2024–25, 24 February 2025. The audit office is independent of the ATO. Its auditors describe the real-time analytics component and find that the ATO did not document the weights behind it. They do not test its revenue effect."", ""url"": ""https://web.archive.org/web/20261014065948/https://www.anao.gov.au/work/performance-audit/governance-of-artificial-intelligence-the-australian-taxation-office""}]","{""kind"": ""Official documents""}" AIBD-0732,2017-01-01,event,year,reported,Evaluators for Energy Trust of Oregon confirm that Nest Seasonal Savings saved about 18 therms per enrolled thermostat in winter 2016-17,"In a randomized pilot, Nest Seasonal Savings cut heating runtime by 4.75%, about 18 therms per opted-in thermostat over winter 2016-17. Summer cooling savings were 4 kWh.",*Nest Seasonal Savings* trims winter heating in Oregon homes,about 18 therms per opted-in thermostat,,claimed,The evaluator's account,randomized trial,high,aggregate result,false,Nest Seasonal Savings (setpoint and schedule optimization service for Nest thermostats),,Nest Labs,United States,"Energy Trust of Oregon with CLEAResult and Nest, offered to existing Nest owners",United States,"Oregon, United States",United States,Energy (residential efficiency),,Energy and infrastructure,Planning and control,Consumers,7,High income,environmental_impact,Environmental impacts,1,3,3,,3,2,,personal; societal,1,0,A control group of Nest thermostats from the same population,Baseline heating runtime of 596 hours,"28.3 hours less, a 4.75% reduction",measured,scaled,2019-04-25,shown,https://www.e3tnw.org/nest-public-downloads/seasonal_savings_white_paper.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Apex Analytics LLC and Demand Side Analytics. Energy Trust of Oregon Nest Thermostat Seasonal Savings Pilot Evaluation. 22 November 2017."", ""url"": ""https://www.energytrust.org/documents/nest-thermostat-seasonal-savings-pilot-evaluation/""}]","{""kind"": ""Official documents""}" AIBD-0882,2015-12-01,source publication,month,reported,Evaluators at Oak Ridge National Laboratory find that socially driven HVAC control raised comfort for 83% of occupants in GSA's Phoenix courthouse,"At the federal courthouse in Phoenix, 83% of occupants reported more thermal comfort with socially driven HVAC control. Calls to maintenance about temperature fell 59%.",Occupants vote to adjust *socially driven HVAC* to their comfort,83% of occupants reported more thermal comfort,,claimed,The evaluator's account,before and after,very low,aggregate result,false,Socially driven HVAC optimization software,,"Building Robotics, LLC (named in the evaluation as the technology provider)",United States,"U.S. General Services Administration (Federal Building and U.S. Courthouse, Phoenix)",United States,"Phoenix, Arizona, United States",United States,Building energy management,,Energy and infrastructure,Planning and control,Public servants,7,High income,operational_improvement,Customer personalization,1,3,4,2,3,4,0.350,personal; organizational; societal,2,0,The building's previous fixed-range temperature control,,83% of occupants reported being more satisfied with their thermal environment,measured,unknown,,shown,https://www.gsa.gov/system/files/GPG025-Socially%20Driven%20HVAC-Findings-508R.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""U.S. General Services Administration, Green Proving Ground. GPG-025: Socially Driven HVAC for Personal Control, findings, December 2015. GSA's summary of the measurement and verification that it commissioned from Oak Ridge National Laboratory in its Phoenix courthouse."", ""url"": ""https://www.gsa.gov/system/files/GPG025-Socially%20Driven%20HVAC-Findings-508R.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Howett D, Rizy CG, Wolfe AK (Oak Ridge National Laboratory). Socially Driven HVAC Optimization: Federal Building and US Courthouse, Phoenix, Arizona. Prepared for the US General Services Administration, August 2015 (archived copy). The evaluators' full report adds the method, the occupant surveys and their own description of how the system learns."", ""url"": ""https://web.archive.org/web/20250430095420id_/https://www.gsa.gov/system/files/Final_Socially_Driven_HVAC_Report_-_508_Compliant.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0731,2015-01-01,event,year,reported,"Evaluators for Energy Trust of Oregon find that self-installed Nest thermostats cut gas heating use about 6%, while Honeywell's Lyric raised it","In Energy Trust of Oregon's 2014-15 pilot, gas-heated homes that self-installed a Nest saved about 34 therms a year, 6% of heating. Homes given Honeywell's Lyric used 4-5% more.",*Nest* thermostats cut gas heating use in Oregon homes by about 6%,Nest homes saved about 34 therms a year,,claimed,The evaluator's account,,very low,aggregate result,false,Nest Learning Thermostat (second generation),,Nest Labs,United States,"Energy Trust of Oregon pilot with NW Natural customers, run by CLEAResult (self-installed thermostats in homes with gas furnaces)",United States,"Oregon, United States",United States,Energy (residential efficiency),,Energy and infrastructure,Planning and control,Consumers,7,High income,environmental_impact,Environmental impacts,1,3,3,1,3,,,personal; societal,1,0,"About 1,000 randomly selected NW Natural customers who were not contacted",,"About 34 therms a year saved, 6.0% of heating usage",measured,scaled,2016-04-06,shown,https://www.energytrust.org/documents/gas-smart-thermostat-pilot-evaluation-report/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Apex Analytics LLC. Energy Trust of Oregon Smart Thermostat Pilot Evaluation. Prepared for Energy Trust of Oregon, 1 March 2016."", ""url"": ""https://www.energytrust.org/documents/gas-smart-thermostat-pilot-evaluation-report/""}]","{""kind"": ""Official documents""}" +AIBD-0731,2015-01-01,event,year,reported,"Evaluators for Energy Trust of Oregon find that self-installed Nest thermostats cut gas heating use about 6%, while Honeywell's Lyric raised it","In Energy Trust of Oregon's 2014-15 pilot, gas-heated homes that self-installed a Nest saved about 34 therms a year, 6% of heating. Homes given Honeywell's Lyric used 4-5% more.",*Nest* thermostats cut gas heating use in Oregon homes by about 6%,Nest homes saved about 34 therms a year,,claimed,The evaluator's account,controlled comparison,low,aggregate result,false,Nest Learning Thermostat (second generation),,Nest Labs,United States,"Energy Trust of Oregon pilot with NW Natural customers, run by CLEAResult (self-installed thermostats in homes with gas furnaces)",United States,"Oregon, United States",United States,Energy (residential efficiency),,Energy and infrastructure,Planning and control,Consumers,7,High income,environmental_impact,Environmental impacts,1,3,3,1,3,,,personal; societal,1,0,"About 1,000 randomly selected NW Natural customers who were not contacted",,"About 34 therms a year saved, 6.0% of heating usage",measured,scaled,2016-04-06,shown,https://www.energytrust.org/documents/gas-smart-thermostat-pilot-evaluation-report/,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Apex Analytics LLC. Energy Trust of Oregon Smart Thermostat Pilot Evaluation. Prepared for Energy Trust of Oregon, 1 March 2016."", ""url"": ""https://www.energytrust.org/documents/gas-smart-thermostat-pilot-evaluation-report/""}]","{""kind"": ""Official documents""}" AIBD-0729,2014-01-01,event,year,reported,"Evaluators for Energy Trust of Oregon find that Nest thermostats cut heat-pump homes' electricity use by 781 kWh a year, 12% of heating load","In Energy Trust of Oregon's 2013-14 pilot, homes given a Nest thermostat to run their heat pumps used 781 kWh a year less, 4.7% of their electricity, than comparison homes.",*Nest* thermostats trim electricity use in Oregon heat-pump homes,781 kWh per year,,claimed,The evaluator's account,controlled comparison,low,aggregate result,false,Nest Learning Thermostat (heat pump control),,Nest Labs,United States,"Energy Trust of Oregon pilot, installed by CLEAResult in 185 heat-pump-heated homes",United States,"Oregon (I-5 corridor), United States",United States,Energy (residential efficiency),,Energy and infrastructure,Planning and control,Consumers,7,High income,environmental_impact,Environmental impacts,1,3,3,4,3,,,personal; societal,1,0,299 comparison homes from the same pool without a Nest,,"781 kWh a year less, 4.7% of total electric usage",measured,withdrawn,2026-01-01,shown,https://www.energytrust.org/wp-content/uploads/2016/12/Nest_Pilot_Study_Evaluation_wSR.pdf,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Evaluation report"", ""value"": ""Apex Analytics LLC. Energy Trust of Oregon Nest Thermostat Heat Pump Control Pilot Evaluation. Prepared for Energy Trust of Oregon, 10 October 2014."", ""url"": ""https://www.energytrust.org/wp-content/uploads/2016/12/Nest_Pilot_Study_Evaluation_wSR.pdf""}, {""label"": ""Follow-up"", ""value"": ""Energy Trust of Oregon announced on 9 December 2025 that from 1 January 2026 it would discontinue its smart thermostat incentive for Oregon homes with electric heating."", ""url"": ""https://insider.energytrust.org/?p=20133""}]","{""kind"": ""Withdrawals""}" AIBD-1164,2026-08-01,source publication,month,tested,"Assessors for India's HTAIn estimate that DeepCxR saves 51,330 rupees per 1,000 people screened for TB but classifies fewer cases correctly","Assessors for India's HTAIn estimated that DeepCxR saves 51,330 rupees per 1,000 people screened for TB against radiologists, with specificity of 62.6% against 89%.",*DeepCxR* would cut TB screening costs at some loss of accuracy,"51,330 rupees per 1,000 people screened",,claimed,The evaluator's account,no comparison,very low,aggregate result,false,DeepCxR (AI-assisted chest X-ray interpretation for TB screening),,"Indian Council of Medical Research (ICMR), Department of Health Research (DHR) and Institute for Plasma Research (IPR), Department of Atomic Energy, Government of India",India,"India's National TB Elimination Program (NTEP), as the program that HTAIn modeled (no deployment described)",India,India,India,Healthcare,,Health,Images and video,Patients,3,Lower middle income,operational_improvement,Enhanced productivity,,,,,,,,organizational,2,0,"Radiologists' interpretation of digital chest X-rays, modeled for 1,000 adults with presumptive TB","Sensitivity 87%, specificity 89%, cost 299,077 INR per 1,000 screened","85.9% Sensitivity and 62.6% Specificity, cost 247,747 INR per 1,000 screened",assumed,unknown,,not shown,https://htain.dhr.gov.in/images/pdf/Policy_Briefs/AI-Assisted_Chest_X-Ray_for_Pulmonary_Tuberculosis.pdf,"51,330 INR","Modeled cost saving per 1,000 individuals screened for TB, against radiologist interpretation",True,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Assessment report"", ""value"": ""Indian Institute of Public Health Gandhinagar, Resource Hub for Health Technology Assessment in India (HTAIn), Department of Health Research, Ministry of Health and Family Welfare. Health Technology Assessment of AI-Assisted (DeepCXR) Chest X-Ray Interpretation for Pulmonary Tuberculosis Screening. Policy brief, August 2026. Conclusion, p. 2. DHR, which runs HTAIn, co-developed DeepCxR."", ""url"": ""https://htain.dhr.gov.in/images/pdf/Policy_Briefs/AI-Assisted_Chest_X-Ray_for_Pulmonary_Tuberculosis.pdf""}]","{""kind"": ""Official documents""}" AIBD-1059,2025-12-04,source publication,day,mixed,NPL finds for the Home Office that the Police National Database's Cognitec algorithm ranked the right person first in 3510 of 3513 test searches,Police use the Cognitec algorithm on the Police National Database. NPL found that it ranked the right person first in 3510 of 3513 test searches. False matches varied by demographic group.,*FaceVACS-DBScan ID* ranks the right face first in 3510 of 3513 tests,the right person first in 3510 of 3513 test searches,,claimed,The evaluator's account,no comparison,very low,aggregate result,false,"Cognitec FaceVACS-DBScan ID, the retrospective facial recognition algorithm of the Police National Database",v5.5,Cognitec Systems,Germany,"Home Office, which provides the Police National Database with the National Police Chiefs' Council, for searches by police forces and law enforcement agencies",United Kingdom,United Kingdom,United Kingdom,Policing,,Safety and rescue,Images and video,Public servants,16,High income,physical_health,Crime reduction or prevention,,3,4,2,2,5,,organizational,3,0,,,,,challenged,2025-12-04,shown,https://www.cognitec.com/technology.html,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Evaluation report"", ""value"": ""National Physical Laboratory (NPL), Data Science and AI Department. Report 4153: Facial Recognition Technologies: Accuracy and Equitability Evaluation of Cognitec FaceVACS-DBScan ID v5.5. Report for the Home Office and the Office of the Policing Chief Scientific Adviser, October 2025, published on GOV.UK on 4 December 2025. Section 1.2.1, Recognition accuracy."", ""url"": ""https://assets.publishing.service.gov.uk/media/693002a4cdec734f4dff4149/1a_Cognitec_NPL_Equitability_Report_October_25.pdf""}, {""label"": ""Evaluation report"", ""value"": ""National Physical Laboratory (NPL), Data Science and AI Department. Report 4153: Facial Recognition Technologies: Accuracy and Equitability Evaluation of Cognitec FaceVACS-DBScan ID v5.5. Report for the Home Office and the Office of the Policing Chief Scientific Adviser, October 2025, published on GOV.UK on 4 December 2025. Section 1.2.2, Demographic variation in TPIR and FPIR, at a face-match threshold of 0.8."", ""url"": ""https://assets.publishing.service.gov.uk/media/693002a4cdec734f4dff4149/1a_Cognitec_NPL_Equitability_Report_October_25.pdf""}, {""label"": ""Follow-up"", ""value"": ""The Home Office reports that it acted on the findings. It procured a new algorithm and had it tested independently, and it asked the police inspectorate, HMICFRS, to conduct an inspection."", ""url"": ""https://www.gov.uk/government/publications/facial-recognition-technology-tests-national-physical-laboratory/summary-accuracy-and-equitability-evaluation-of-cognitec-facevacs-dbscan-id-v55""}]","{""kind"": ""Challenges""}" AIBD-0412,2025-06-01,source information,month,tested,GeoGPT: AI that monitors marine ecosystems to protect them,"Zhejiang Lab reports that its GeoGPT platform identifies deep-sea species with 81.3% accuracy. It takes 40 milliseconds per image, compared with 40 seconds manually.",*GeoGPT* identifies deep-sea species in images faster than people do,,,claimed,The makers' claim,no comparison,very low,aggregate result,false,GeoGPT,,Zhejiang Lab,China,,,,,Environment and climate,,Environment,Images and video,Researchers,14,,operational_improvement,Enhanced productivity,,,,,,,,organizational; societal,2,0,manual identification of deep-sea species in images,40 seconds per image by a person,40 milliseconds per image by GeoGPT,assumed,challenged,2026-07-25,shown,https://www.itu.int/dms_pub/itu-t/opb/ai4g/T-AI4G-AI4GOOD-2025-5-PDF-E.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""In July 2026 Cleverley and colleagues reported that the International Union of Geological Sciences had suspended Deep-time Digital Earth, the program behind GeoGPT, in early 2025. They found that its claims of openness and neutrality could not be verified."", ""url"": ""https://www.journalofgeoethics.eu/index.php/jgsg/article/view/119""}]","{""kind"": ""Challenges""}" -AIBD-0520,2025-06-01,source information,month,intended,AI that detects drowsy drivers in real time to improve road safety,Catalink built an AI drowsiness detector for drivers. In July 2026 its researchers reported that their models generalized poorly. Phone versions fell to near chance-level accuracy.,Drowsiness models for drivers perform poorly in their makers' own tests,,,claimed,The makers' claim,,not rated,program,false,Catalink AI drowsiness-detection system,,Catalink,Cyprus,,,,,Accessibility,,Accessibility,Images and video,Drivers and travelers,3; 11,,physical_health,Hazard reduction,,,,,,,,personal; organizational,3,0,Current driver monitoring systems built on facial landmark models,struggle to accurately detect drowsiness leading to false negatives,,assumed,challenged,2026-07-23,shown,https://www.itu.int/dms_pub/itu-t/opb/ai4g/T-AI4G-AI4GOOD-2025-5-PDF-E.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""In July 2026 the Catalink team reported that the models that they tested generalized poorly across datasets and that the versions converted for phones fell to near chance-level accuracy. The work remained offline research."", ""url"": ""https://zenodo.org/records/21512007""}]","{""kind"": ""Challenges""}" -AIBD-0597,2025-05-01,event,day,unclear,Virginia ended its pilot of AI for reviewing regulations without publishing results,In 2025 Virginia ran a pilot of Vulcan Justinian to analyze state regulations. It ended the pilot in 2026. Virginia has published no estimate of how much the AI contributed to the claimed savings.,Virginia tested AI to help agencies review regulations,,,claimed,The makers' claim,,not rated,program,false,Vulcan Justinian,,Vulcan Technologies,,Virginia Office of Regulatory Management and executive agencies,,"Virginia, United States",United States,Public administration,Regulatory review,Public sector,Text and language,Public servants,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,5,0,,,,,withdrawn,2026-06-30,shown,https://vulcan.ai/,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Regulatory document"", ""value"": ""Abigail D. Spanberger, Governor of Virginia, Executive Order No. 17 (2026), Development and Review of State Agency Regulations, 30 June 2026. The order comes from the state executive branch that ran the pilot, under a new administration. It ends the pilot and its guidance and contains no finding on its results."", ""url"": ""https://www.governor.virginia.gov/media/governorvirginiagov/governor-of-virginia/pdf/eo/eo-17-development-and-review-of-state-agency-regulations.pdf""}, {""label"": ""Follow-up"", ""value"": ""Governor Abigail Spanberger signed Executive Order 17 on 30 June 2026. It rescinds Executive Order 51 and all guidance on the AI Regulatory Reduction Pilot from 1 July 2026. It refers to the Office of Regulatory Management as a former office."", ""url"": ""https://www.governor.virginia.gov/media/governorvirginiagov/governor-of-virginia/pdf/eo/eo-17-development-and-review-of-state-agency-regulations.pdf""}]","{""kind"": ""Withdrawals""}" +AIBD-0520,2025-06-01,source information,month,intended,AI that detects drowsy drivers in real time to improve road safety,Catalink built an AI drowsiness detector for drivers. In July 2026 its researchers reported that their models generalized poorly. Phone versions fell to near chance-level accuracy.,Drowsiness models for drivers perform poorly in their makers' own tests,,,claimed,The makers' claim,,very low,program,false,Catalink AI drowsiness-detection system,,Catalink,Cyprus,,,,,Accessibility,,Accessibility,Images and video,Drivers and travelers,3; 11,,physical_health,Hazard reduction,,,,,,,,personal; organizational,3,0,Current driver monitoring systems built on facial landmark models,struggle to accurately detect drowsiness leading to false negatives,,assumed,challenged,2026-07-23,shown,https://www.itu.int/dms_pub/itu-t/opb/ai4g/T-AI4G-AI4GOOD-2025-5-PDF-E.pdf,,,,[],Querist,The benefit was later disputed.,"[{""label"": ""Follow-up"", ""value"": ""In July 2026 the Catalink team reported that the models that they tested generalized poorly across datasets and that the versions converted for phones fell to near chance-level accuracy. The work remained offline research."", ""url"": ""https://zenodo.org/records/21512007""}]","{""kind"": ""Challenges""}" +AIBD-0597,2025-05-01,event,day,unclear,Virginia ended its pilot of AI for reviewing regulations without publishing results,In 2025 Virginia ran a pilot of Vulcan Justinian to analyze state regulations. It ended the pilot in 2026. Virginia has published no estimate of how much the AI contributed to the claimed savings.,Virginia tested AI to help agencies review regulations,,,claimed,The makers' claim,,very low,program,false,Vulcan Justinian,,Vulcan Technologies,,Virginia Office of Regulatory Management and executive agencies,,"Virginia, United States",United States,Public administration,Regulatory review,Public sector,Text and language,Public servants,16,High income,operational_improvement,Enhanced productivity,,,,,,,,organizational,5,0,,,,,withdrawn,2026-06-30,shown,https://vulcan.ai/,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Regulatory document"", ""value"": ""Abigail D. Spanberger, Governor of Virginia, Executive Order No. 17 (2026), Development and Review of State Agency Regulations, 30 June 2026. The order comes from the state executive branch that ran the pilot, under a new administration. It ends the pilot and its guidance and contains no finding on its results."", ""url"": ""https://www.governor.virginia.gov/media/governorvirginiagov/governor-of-virginia/pdf/eo/eo-17-development-and-review-of-state-agency-regulations.pdf""}, {""label"": ""Follow-up"", ""value"": ""Governor Abigail Spanberger signed Executive Order 17 on 30 June 2026. It rescinds Executive Order 51 and all guidance on the AI Regulatory Reduction Pilot from 1 July 2026. It refers to the Office of Regulatory Management as a former office."", ""url"": ""https://www.governor.virginia.gov/media/governorvirginiagov/governor-of-virginia/pdf/eo/eo-17-development-and-review-of-state-agency-regulations.pdf""}]","{""kind"": ""Withdrawals""}" AIBD-0278,2025-01-16,source publication,day,tested,"Researchers synthesize MatterGen's TaCr2O6, which a later analysis shows to have been known since 1971","Researchers synthesized TaCr2O6, which MatterGen generated. They estimated its bulk modulus at 169 GPa against a 200 GPa target. A chemist later found that the compound had been known since 1971.",MatterGen generates a material later found to be known since 1971,estimated its bulk modulus at 169 GPa against a 200 GPa target,,claimed,The makers' claim,tested against a standard,very low,benefit created,false,MatterGen,,Microsoft,United States,"Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences (synthesis)",China,"Cambridge, UK (model); Shenzhen, China (synthesis)",United Kingdom | China,Materials science,Materials researchers,Science,Prediction from data,Researchers,9,Upper middle income; High income,knowledge_discovery,,,1,,,,1,,personal; organizational; societal,4,0,"previous generative models for crystal structures (DiffCSP, CDVAE and others) on the same benchmark",not stated as a figure in the text (previous generative models),"more than twice as likely to be SUN (stable, unique and new); the synthesized TaCr2O6 measured at 169 GPa against a 200 GPa design target",measured,reversed,2026-06-12,shown,https://pmc.ncbi.nlm.nih.gov/articles/PMC11922738/,,,,[],Querist,The benefit was later reversed.,"[{""label"": ""Follow-up"", ""value"": ""In Materials Horizons on 12 June 2026, Mikkel Juelsholt reported that the synthesized disordered TaCr2O6 is identical to Ta1/2Cr1/2O2. That compound was first described in 1971 and sat in MatterGen's training data, so the material of the proof of concept was not new."", ""url"": ""https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:42007796%20AND%20SRC:MED&resultType=core&format=json""}]","{""kind"": ""Challenges""}" AIBD-0925,2024-01-01,event,year,reported,"Guidehouse, the reviewer that Meta proposed, verifies that the Variance Reduction System met targets for housing ads under Meta's settlement with the Justice Department","Guidehouse verified that Meta met its targets for housing ads from January to April 2024. Of ads with at least 300 impressions, 94.2% had a sex variance of 10% or less, against a required 90.2%.",*Variance Reduction System* keeps Meta's housing ads to agreed limits,"94.2% had a sex variance of 10% or less, against a required 90.2%",,claimed,The evaluator's account,tested against a standard,low,aggregate result,false,Variance Reduction System (VRS),,"Meta Platforms, Inc.",United States,"Meta Platforms, Inc. (housing ads on Facebook, Instagram and Messenger)",United States,United States,United States,Housing advertising,,Work and industry,Planning and control,The public,10,High income,opportunity_access,Housing access,,,3,3,3,5,,societal,3,0,The VRS Compliance Metrics that the Department of Justice and Meta agreed and the court adopted,"Required coverage of 90.2% (sex, variance of 10% or less, ads with at least 300 impressions)",94.2%,measured,persisted,2024-10-30,shown,https://about.fb.com/wp-content/uploads/2023/01/Toward_fairness_in_personalized_ads.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Guidehouse Inc., independent third-party reviewer under the Settlement Agreement in United States v. Meta Platforms, Inc., No. 22-Civ-5187 (S.D.N.Y.). Meta VRS Compliance Metrics Verification, 28 June 2024. I. Executive Summary, p. 1, with Tables 1 and 2, p. 3. Meta proposed Guidehouse as reviewer with the consent of the Department of Justice. Published by the U.S. Department of Justice, Civil Rights Division."", ""url"": ""https://www.justice.gov/crt/media/1362086/dl""}, {""label"": ""Evaluation report"", ""value"": ""Guidehouse Inc. Meta VRS Compliance Metrics Verification, 30 October 2024, for the reporting period of 1 May to 31 August 2024. I. Executive Summary, p. 1. Published by the U.S. Department of Justice, Civil Rights Division."", ""url"": ""https://www.justice.gov/crt/media/1385866/dl""}]","{""kind"": ""Official documents""}" AIBD-0036,2023-12-27,event,day,reported,"An automated classifier on a glider detects right whale calls, which prompt a voluntary NOAA Slow Zone east of Portland, Maine","The Low-Frequency Detection and Classification System on a glider detected right whales east of Portland, Maine, on 27 December 2023. NOAA declared a voluntary Slow Zone.",NOAA asks mariners to slow after glider software detects right whales,route around the area or transit at 10 knots or less,,claimed,The deployer's account,no comparison,very low,protective response,false,DMON instrument running the Low-Frequency Detection and Classification System (LFDCS) on a Slocum glider,,Woods Hole Oceanographic Institution (Mark Baumgartner),United States,Woods Hole Oceanographic Institution and NOAA Northeast Fisheries Science Center; NOAA Fisheries issues Slow Zones,United States,"Gulf of Maine east of Portland, United States",United States,Marine conservation / shipping,Endangered-species vessel-strike mitigation,Environment,Speech and sound,,14,High income,animal_health; environmental_impact,Environmental impacts,,,2,,5,1,,societal,6,2,aerial surveillance,,more slowdowns than aerial surveillance by a factor of two to five during fall and summer,measured,scaled,2025-01-21,shown,https://dcs.whoi.edu/lfdcs_manual/lfdcs_manual.html,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""NOAA Fisheries (29 December 2023). New Right Whale Slow Zone: East of Portland, ME - Effective Through 1/11. GovDelivery bulletin."", ""url"": ""https://content.govdelivery.com/accounts/USNOAAFISHERIES/bulletins/3824b02""}, {""label"": ""Regulatory document"", ""value"": ""NOAA Fisheries. Reducing Vessel Strikes to North Atlantic Right Whales (Slow Zones policy page)."", ""url"": ""https://www.fisheries.noaa.gov/national/endangered-species-conservation/reducing-vessel-strikes-north-atlantic-right-whales""}]","{""kind"": ""Official documents""}" AIBD-0482,2023-11-05,source publication,day,tested,"Industrial quality inspection with AI, based on the IndustryGPT multimodal large model","SmartMore reports that, in one classification test, training set up by IndustryGPT reached 100% accuracy in 5:27, compared with 50% in 9:31 for manual training.",IndustryGPT beats manual training in an industrial inspection test,,,claimed,The makers' claim,tested against a standard,very low,aggregate result,false,IndustryGPT and SMore ViMo,,SmartMore,China,,,,,Manufacturing,,Work and industry,Text and language; Images and video,,9,,operational_improvement,Enhanced productivity,,,1,,,,,organizational,3,0,Manual model training for the same classification detection task,"9 minutes 31 seconds of training, 50% accuracy and 66.7% recall with manual training","5 minutes 27 seconds of training, 100% accuracy and 100% recall with the solution chosen by IndustryGPT",measured,scaled,2026-03-16,shown,https://www.itu.int/dms_pub/itu-t/opb/ai4g/T-AI4G-AI4GOOD-2025-5-PDF-E.pdf,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Regulatory document"", ""value"": ""SmartMore Inc., Application Proof filed with The Stock Exchange of Hong Kong, March 2026. The company wrote this draft of its listing document. The exchange and the Securities and Futures Commission take no responsibility for its contents. In it the company reports company-wide figures without isolating the classification example."", ""url"": ""https://www1.hkexnews.hk/app/sehk/2026/108311/documents/sehk26031600232.pdf""}]","{""kind"": ""Official documents""}" -AIBD-0353,2023-02-06,event,day,unclear,"In HOT's activation for the Türkiye earthquake of February 2023, AI-assisted RapiD mapping was up to 12.4% slower. It was disabled. OSM data still reached AKUT and AFAD","After the 2023 Türkiye earthquake, researchers found that contributors using RapiD mapped tasks up to 12.4% slower. RapiD is an AI-assisted editor. It was then disabled.",Türkiye quake: Mappers using AI-assisted RapiD were up to 12.4% slower,contributors using RapiD mapped tasks up to 12.4% slower,,claimed,The deployer's account,,not rated,aggregate result,false,"RapiD editor (AI-assisted mapping, maintained by Meta) used on HOT Tasking Manager projects","MapWithAI AI-predicted buildings and roads, offered in the Rapid editor (maintained by Meta) and in the JOSM MapWithAI plugin","Meta, which maintains the Rapid editor",,Humanitarian OpenStreetMap Team with Yer Çizenler (activation lead),United States,Türkiye (Kahramanmaraş earthquakes) and northern Syria,Turkey | Syria,Humanitarian mapping / disaster response,Volunteer humanitarian mapping,Humanitarian,Images and video,Other workers,11,Low income; Upper middle income,operational_improvement,"Faster data management, decision-making",,,,,,,,personal; organizational; societal,5,1,"Tasking Manager contributors who did not use AI-assisted mapping, and the same 1,650 contributors in the weeks before the response",,Contributors using Rapid mapped tasks up to 12.4% slower,measured,withdrawn,,shown,https://wiki.openstreetmap.org/wiki/Rapid,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Follow-up"", ""value"": ""By the wiki's account, validators' complaints led HOT to disable Rapid on all activation projects, at a date that it does not give. The archived settings of the 11 Türkiye projects still list Rapid as a mapping editor, so the extent of the withdrawal is uncertain."", ""url"": ""https://wiki.openstreetmap.org/wiki/2023_Turkey_Earthquakes""}]","{""kind"": ""Withdrawals""}" +AIBD-0353,2023-02-06,event,day,unclear,"In HOT's activation for the Türkiye earthquake of February 2023, AI-assisted RapiD mapping was up to 12.4% slower. It was disabled. OSM data still reached AKUT and AFAD","After the 2023 Türkiye earthquake, researchers found that contributors using RapiD mapped tasks up to 12.4% slower. RapiD is an AI-assisted editor. It was then disabled.",Türkiye quake: Mappers using AI-assisted RapiD were up to 12.4% slower,contributors using RapiD mapped tasks up to 12.4% slower,,claimed,The deployer's account,,very low,aggregate result,false,"RapiD editor (AI-assisted mapping, maintained by Meta) used on HOT Tasking Manager projects","MapWithAI AI-predicted buildings and roads, offered in the Rapid editor (maintained by Meta) and in the JOSM MapWithAI plugin","Meta, which maintains the Rapid editor",,Humanitarian OpenStreetMap Team with Yer Çizenler (activation lead),United States,Türkiye (Kahramanmaraş earthquakes) and northern Syria,Turkey | Syria,Humanitarian mapping / disaster response,Volunteer humanitarian mapping,Humanitarian,Images and video,Other workers,11,Low income; Upper middle income,operational_improvement,"Faster data management, decision-making",,,,,,,,personal; organizational; societal,5,1,"Tasking Manager contributors who did not use AI-assisted mapping, and the same 1,650 contributors in the weeks before the response",,Contributors using Rapid mapped tasks up to 12.4% slower,measured,withdrawn,,shown,https://wiki.openstreetmap.org/wiki/Rapid,,,,[],Querist,The system was later withdrawn.,"[{""label"": ""Follow-up"", ""value"": ""By the wiki's account, validators' complaints led HOT to disable Rapid on all activation projects, at a date that it does not give. The archived settings of the 11 Türkiye projects still list Rapid as a mapping editor, so the extent of the withdrawal is uncertain."", ""url"": ""https://wiki.openstreetmap.org/wiki/2023_Turkey_Earthquakes""}]","{""kind"": ""Withdrawals""}" AIBD-0700,2023-01-01,event,year,reported,An evaluation for UNDP shows that an AI application widened Liberia's iVerify monitoring to hate speech on radio in the 2023 elections,"An evaluation for UNDP shows that an AI application, integrated with iVerify, captured hate speech on radio. It widened Liberia's monitoring in the 2023 elections.",AI captures hate speech on radio for Liberia's *iVerify* fact-checkers,,,claimed,The evaluator's account,no comparison,very low,program,false,"AI application for automated capture of hate speech on radio, integrated with UNDP's iVerify platform (application not named)",,Not named in the evaluation (Peacebuilding Fund project on a peaceful electoral environment),,"Local Voices Liberia (iVerify operator), with the National Elections Commission and UNDP",Liberia,Liberia,Liberia,Electoral integrity / hate speech monitoring,,Public sector,Speech and sound,Other workers,16,Low income,social_improvement,,,,,,3,,,organizational; societal,1,0,,,,,persisted,2024-04-08,not shown,https://www.unevaluation.org/get-download-file/17401,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Nelson, S. Midterm Evaluation Completion, Liberian Electoral Support Project (LESP): Final Report. Independent evaluation commissioned by UNDP Liberia, 9 April 2025. Published in the UNDP and UN Evaluation Group repositories."", ""url"": ""https://www.unevaluation.org/get-download-file/17401""}, {""label"": ""Evaluation report"", ""value"": ""Nelson, S. Midterm Evaluation Completion, Liberian Electoral Support Project (LESP): Final Report. Independent evaluation commissioned by UNDP Liberia, 9 April 2025. Published in the UNDP and UN Evaluation Group repositories."", ""url"": ""https://www.unevaluation.org/get-download-file/17401""}]","{""kind"": ""Official documents""}" AIBD-0701,2023-01-01,event,year,reported,An evaluation for UNDP shows that Sierra Leone's iVerify team processed 265 fact-checking requests during the tense 2023 election period,An evaluation for UNDP shows that Sierra Leone's iVerify team processed 265 fact-checking requests around the 2023 elections. The evaluator does not mention AI.,Sierra Leone's *iVerify* team handles 265 fact-checking requests,265 fact-checking requests,,claimed,The evaluator's account,no comparison,very low,program,false,iVerify (UNDP fact-checking platform),,UNDP (Chief Digital Office and the Brussels-based Task Force on Electoral Assistance),,"Sierra Leone Association of Journalists, Independent Radio Network and BBC Media Action, with UNDP's elections project",Sierra Leone,Sierra Leone,Sierra Leone,Electoral integrity / fact-checking,,Public sector,Text and language,Other workers,16,Low income,social_improvement,Fact checking,,,4,2,3,,,societal,5,1,,,,,persisted,2024-09-23,shown,https://web.archive.org/web/20260713191012/https://www.undp.org/digital/stories/ai-powered-fact-checking-tool-iverify-piloted-during-zambia-election-shows-global-promise,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Robertson, L. Evaluation of Electoral Assistance Project 'Fostering Peaceful, Credible and Inclusive Elections in Sierra Leone 2022-2023'. Independent evaluation commissioned by the UNDP Sierra Leone Country Office, June 2024. Published in the UNDP and UN Evaluation Group repositories."", ""url"": ""https://www.unevaluation.org/get-download-file/4321""}, {""label"": ""Evaluation report"", ""value"": ""UNDP Independent Evaluation Office. Evaluation of the UNDP Support to Digitalization of Public Services. November 2023. The report is an independent publication of the office. Section on electoral support."", ""url"": ""https://www.unevaluation.org/get-download-file/4444""}, {""label"": ""Assessment report"", ""value"": ""European Union Election Observation Mission Sierra Leone 2023, Final Report: General Elections 24 June 2023, published by the EEAS on 10 October 2023, section XII (Social media and digital rights)."", ""url"": ""https://www.eeas.europa.eu/sites/default/files/documents/2023/EU%20EOM%20SLE%202023%20Final%20Report.pdf""}, {""label"": ""Evaluation report"", ""value"": ""Dr Edwin Okul, independent consultant commissioned by UNDP and UNICEF Sierra Leone (2025). Final Evaluation of the Peacebuilding (PBF) Project: Promote the creation of an enabling environment for the conduct of peaceful elections and the strengthening of social cohesion in Sierra Leone, p. 38 (PDF p. 47), Output 2.4, Indicator 2.4.1."", ""url"": ""https://erc.undp.org/evaluation/documents/download/24794""}]","{""kind"": ""Official documents""}" AIBD-0923,2023-01-01,event,year,reported,An evaluation for UNDP shows that Mozambique's election commission received findings from eMonitor+ on misinformation for the 2023 local elections,An evaluation for UNDP shows that findings from eMonitor+ went to Mozambique's National Elections Commission to help it address false information around the 2023 local elections.,*eMonitor+* findings reach Mozambique's election commission in 2023,,,claimed,The evaluator's account,no comparison,very low,program,false,"eMonitor+, UNDP's platform for monitoring digital media in elections",,"UNDP Regional Bureau for Arab States (developed in Tunisia, according to the evaluation)",,"UNDP Mozambique's Enhancing Democracy and Electoral Processes (EDEP) project, staffed by UN Volunteers, for Mozambique's National Elections Commission (CNE)",Mozambique,Mozambique (national),Mozambique,Electoral integrity / information monitoring,,Public sector,Text and language,Public servants,16,Low income,social_improvement,Fact checking,,1,3,,3,,,organizational; societal,2,0,,,,,persisted,2024-10-09,shown,https://web.archive.org/web/20260112231245/https://www.undp.org/arab-states/press-releases/election-commission-uses-artificial-intelligent-enhance-womens-participation-electoral-processes,,,,[],Querist,The sources include an official document about the benefit.,"[{""label"": ""Evaluation report"", ""value"": ""Robertson, L. Mid-Term Evaluation Report: Enhancing Democracy and Electoral Processes in Mozambique. Independent evaluation commissioned by UNDP Mozambique, 9 October 2024, p. 17 (Output 1). Published in the UNDP Evaluation Resource Center."", ""url"": ""https://erc.undp.org/evaluation/documents/download/25436""}, {""label"": ""Evaluation report"", ""value"": ""Robertson, L. Mid-Term Evaluation Report: Enhancing Democracy and Electoral Processes in Mozambique. Independent evaluation commissioned by UNDP Mozambique, 9 October 2024, p. 20 (Result 2.5). Published in the UNDP Evaluation Resource Center."", ""url"": ""https://erc.undp.org/evaluation/documents/download/25436""}]","{""kind"": ""Official documents""}"