Part 3: retrain on legally-grounded data, document the more-data-is-not-better regression and the macro-F1 fix
dd318c1 verified | text,label | |
| We're planning to use machine learning to a bank's AI system calculating a loan applicant's credit score to approve or reject a mortgage.,high_risk | |
| Our client wants to deploy an AI tool to a shopping mall's live facial recognition system feeding continuous passer-by identification to police with no judicial authorisation.,prohibited | |
| A service that identifies unnamed individuals in old CCTV recordings by matching their faces to a reference database.,high_risk | |
| A startup is building a product that uses AI to software generating draft small-claims rulings from structured case data and standard legal templates.,high_risk | |
| Our client wants to deploy an AI tool to software grading student exams and quizzes that count toward a final course evaluation.,high_risk | |
| "We're planning to use machine learning to an insurer's AI reviewing life-insurance applications using age, health status and lifestyle to decide terms.",high_risk | |
| "Proposed use case: an AI system designed to an insurer's AI reviewing life-insurance applications using age, health status and lifestyle to decide terms.",high_risk | |
| Our client wants to deploy an AI tool to a school hallway camera system flagging bullying incidents outside of any testing context.,minimal_risk | |
| "We're planning to use machine learning to an AI tool forecasting electricity demand for grid capacity planning, separate from any safety control system.",minimal_risk | |
| We're planning to use machine learning to a music app recommending the next song based on listening history.,minimal_risk | |
| Our client wants to deploy an AI tool to a police-used tool predicting which specific individuals are at risk of becoming victims of trafficking.,high_risk | |
| Software categorising conference attendees by inferred demographic traits based on facial biometric data.,high_risk | |
| AI-based digital forensics software detecting whether CCTV footage submitted as evidence has been altered.,high_risk | |
| "We are developing an AI system that will a retailer running a legal targeted-advertising campaign based on browsing history, using ordinary marketing persuasion.",minimal_risk | |
| A system that detects members of a specific ethnic minority and steers them toward disadvantageous contract terms.,prohibited | |
| We're planning to use machine learning to a toy that uses AI to exploit small children's trust to pressure them into asking parents to buy in-app add-ons.,prohibited | |
| "A job-matching tool that scores and ranks candidates from CVs, generating a 'top 5' shortlist for recruiters.",high_risk | |
| We are developing an AI system that will an AI tool matching apprenticeship applicants to available vocational training slots based on qualifications.,high_risk | |
| A startup is building a product that uses AI to software flagging potential typos in a document before publishing.,minimal_risk | |
| We are developing an AI system that will a vending machine using biometric age estimation to decide whether to sell an age-restricted product.,high_risk | |
| We're planning to use machine learning to an admissions algorithm deciding which applicants get into a university's engineering programme.,high_risk | |
| We are developing an AI system that will software flagging potential typos in a document before publishing.,minimal_risk | |
| We're planning to use machine learning to an AI system classifying 112 emergency calls by urgency and routing responders using natural language processing.,high_risk | |
| We are developing an AI system that will a call-taker assistant that listens for signs of a life-threatening emergency and helps classify call severity.,high_risk | |
| A startup is building a product that uses AI to AI software assisting an arbitration panel in interpreting contract law for a dispute resolution.,high_risk | |
| "We're planning to use machine learning to a tool assigning a 'future offender' risk label to a named person based purely on inferred personality traits, with no verifiable facts.",prohibited | |
| We're planning to use machine learning to a marketing team using AI to generate product photos that look like real studio photography.,limited_risk | |
| Proposed use case: an AI system designed to software scheduling campaign rallies and assigning volunteers to shifts for a political party.,minimal_risk | |
| Our client wants to deploy an AI tool to an insurer's AI pricing individual life-insurance premiums based on predicted health risk.,high_risk | |
| Software using flicker rates below conscious perception embedded in an ad to increase impulse buying.,prohibited | |
| "Our client wants to deploy an AI tool to a pronunciation-practice tool giving students feedback on their accent, not used by teachers to grade them.",minimal_risk | |
| An AI system producing a per-traveller risk score from travel history and watch-list matches for border referral.,high_risk | |
| Proposed use case: an AI system designed to a tool assessing the credibility of witness statements by checking internal consistency against verified facts.,high_risk | |
| "Software monitoring which bills an elected official has sponsored, for public transparency reporting.",minimal_risk | |
| We're planning to use machine learning to an AI tool used to decide employee promotions from automated productivity scoring alone.,high_risk | |
| We are developing an AI system that will a 911/112 dispatch system that automatically prioritises which emergency calls get a first responder sent first.,high_risk | |
| We are developing an AI system that will a marketing tool de-duplicating a mailing list before a newsletter send.,minimal_risk | |
| Proposed use case: an AI system designed to a music app recommending the next song based on listening history.,minimal_risk | |
| Our client wants to deploy an AI tool to software generating a synthetic voiceover that sounds like a real professional narrator for an audiobook.,limited_risk | |
| We're planning to use machine learning to a system that detects members of a specific ethnic minority and steers them toward disadvantageous contract terms.,prohibited | |
| A tool assessing the credibility of witness statements by checking internal consistency against verified facts.,high_risk | |
| Our client wants to deploy an AI tool to law-enforcement software profiling individuals during the detection phase of an ongoing criminal case.,high_risk | |
| Our client wants to deploy an AI tool to software used by a public agency to decide whether to revoke someone's disability support payments.,high_risk | |
| A shift-scheduling AI that ranks warehouse workers by performance and allocates the best shifts accordingly.,high_risk | |
| We are developing an AI system that will an app that scans a person's face to infer their likely religious affiliation for targeted advertising.,prohibited | |
| "We're planning to use machine learning to a plagiarism checker comparing submitted homework against a database of existing text, done after submission.",minimal_risk | |
| We're planning to use machine learning to an AI tool comparing footage from a private security camera against a database to identify a suspect seen on the recording.,high_risk | |
| Proposed use case: an AI system designed to an app estimating a shopper's age bracket from a photo purely to personalise in-store advertising.,minimal_risk | |
| A startup is building a product that uses AI to a cloud provider's load-balancer distributing traffic across internal servers.,minimal_risk | |
| We're planning to use machine learning to software assigning primary-school children to specific classes or streams based on an automated assessment.,high_risk | |
| An AI system on a surveillance tower flagging human presence near a land border to trigger patrol dispatch.,high_risk | |
| Proposed use case: an AI system designed to an internal ticket-routing tool sorting IT support requests by category.,minimal_risk | |
| Our client wants to deploy an AI tool to a government system automatically deciding eligibility for unemployment benefit applications.,high_risk | |
| Our client wants to deploy an AI tool to an automated system determining scholarship eligibility based on an applicant's financial situation.,high_risk | |
| We are developing an AI system that will an AI system deciding whether an applicant qualifies for unemployment benefits and in what amount.,high_risk | |
| Our client wants to deploy an AI tool to an AI learning assistant whose end-of-unit reports on student progress are used by teachers to set final grades.,high_risk | |
| A startup is building a product that uses AI to an automated essay-grading system that decides whether a student passes or fails a course.,high_risk | |
| We're planning to use machine learning to a private security firm's tool labelling named individuals as 'high crime risk' from personality-test results alone.,prohibited | |
| A startup is building a product that uses AI to a music app recommending the next song based on listening history.,minimal_risk | |
| "Proposed use case: an AI system designed to a reoffending-risk model used by a parole board to help decide early release, based on more than just profiling.",high_risk | |
| We're planning to use machine learning to software predicting a risk group's expected annual medical cost to set the health-insurance premium.,high_risk | |
| We are developing an AI system that will an online banking app verifying a customer's identity via facial scan to let them log in.,minimal_risk | |
| Our client wants to deploy an AI tool to an AI system recognising heavy objects on a vulnerable bridge to prevent structural collapse.,high_risk | |
| We're planning to use machine learning to an internal ticket-routing tool sorting IT support requests by category.,minimal_risk | |
| A warehouse system forecasting next month's demand for a product line.,minimal_risk | |
| We are developing an AI system that will a local news site auto-publishing AI-written articles about public interest topics with no editorial review.,limited_risk | |
| We are developing an AI system that will an asylum agency's AI tool assisting officials in examining the eligibility of asylum applications.,high_risk | |
| "Proposed use case: an AI system designed to a retailer running a legal targeted-advertising campaign based on browsing history, using ordinary marketing persuasion.",minimal_risk | |
| "Proposed use case: an AI system designed to a call-centre tool detecting customer frustration from voice tone to route calls, outside any workplace/education context.",high_risk | |
| Our client wants to deploy an AI tool to a shift-scheduling AI that ranks warehouse workers by performance and allocates the best shifts accordingly.,high_risk | |
| We are developing an AI system that will a farm-management tool predicting optimal irrigation timing from soil-moisture sensors.,minimal_risk | |
| "A tool helping a loan applicant understand and fill out a credit application form, separate from the scoring itself.",minimal_risk | |
| Proposed use case: an AI system designed to software matching CCTV facial images from a stadium incident against a database to identify offenders after the fact.,high_risk | |
| We're planning to use machine learning to predictive maintenance software forecasting when a factory machine needs servicing from vibration sensors.,minimal_risk | |
| A toy that uses AI to exploit small children's trust to pressure them into asking parents to buy in-app add-ons.,prohibited | |
| Proposed use case: an AI system designed to a brain-computer interface product that uses subliminal audio cues to push users toward a specific financial choice.,prohibited | |
| We are developing an AI system that will an AI tool enhancing a blurry CCTV image so a human investigator can examine it more clearly.,minimal_risk | |
| A startup is building a product that uses AI to an online store suggesting related products based on a shopper's own past purchases on that same store.,minimal_risk | |
| We're planning to use machine learning to software assisting a visa office in assessing the reliability of evidence submitted with a residence-permit application.,high_risk | |
| A voice assistant that uses manipulative conversational tactics designed to wear down an elderly user's resistance until they agree to a costly subscription.,prohibited | |
| "Proposed use case: an AI system designed to software assisting a judge's sentencing decision using documented case facts alongside a risk score, not personality alone.",high_risk | |
| We are developing an AI system that will a spam filter sorting unwanted emails out of a company inbox.,minimal_risk | |
| We're planning to use machine learning to a retail analytics tool inferring shoppers' political leanings from facial biometric scans in-store.,prohibited | |
| A startup is building a product that uses AI to an app that identifies users with cognitive disabilities and adjusts its interface to make them more likely to accept unfavourable terms.,prohibited | |
| A startup is building a product that uses AI to image-generation software producing photorealistic product renders used in an online catalogue.,limited_risk | |
| A startup is building a product that uses AI to software using flicker rates below conscious perception embedded in an ad to increase impulse buying.,prohibited | |
| An automated system determining scholarship eligibility based on an applicant's financial situation.,high_risk | |
| Our client wants to deploy an AI tool to an AI tool matching apprenticeship applicants to available vocational training slots based on qualifications.,high_risk | |
| We're planning to use machine learning to software scheduling campaign rallies and assigning volunteers to shifts for a political party.,minimal_risk | |
| "We are developing an AI system that will a company harvesting profile pictures from across the internet, untargeted, to train and grow a face-matching product.",prohibited | |
| Our client wants to deploy an AI tool to software estimating shipping costs based on package weight and distance.,minimal_risk | |
| An AI learning assistant whose end-of-unit reports on student progress are used by teachers to set final grades.,high_risk | |
| A police-used tool predicting which specific individuals are at risk of becoming victims of trafficking.,high_risk | |
| "Software predicting which neighbourhood and time window has elevated burglary risk, without linking to a specific person.",minimal_risk | |
| We're planning to use machine learning to maritime AI tracking and detecting persons for border-control interception operations.,high_risk | |
| "We're planning to use machine learning to an outreach tool proactively identifying people who might benefit from preventive healthcare, without deciding eligibility.",minimal_risk | |
| "Our client wants to deploy an AI tool to a job-matching tool that scores and ranks candidates from CVs, generating a 'top 5' shortlist for recruiters.",high_risk | |
| Our client wants to deploy an AI tool to a warehouse system forecasting next month's demand for a product line.,minimal_risk | |
| A startup is building a product that uses AI to software allocating limited home-care hours among elderly applicants based on need.,high_risk | |
| We're planning to use machine learning to a local news site auto-publishing AI-written articles about public interest topics with no editorial review.,limited_risk | |
| "Our client wants to deploy an AI tool to a border-control AI system used to identify individuals during migration processing, beyond simple travel-document checks.",high_risk | |
| An automated tool determining eligibility for subsidised healthcare enrollment.,high_risk | |
| "Proposed use case: an AI system designed to software flagging unusual transaction patterns that may indicate money laundering, based only on amounts and timing, not personal profiles.",minimal_risk | |
| Our client wants to deploy an AI tool to a polygraph-style AI tool used during a police interrogation to assess truthfulness.,high_risk | |
| A startup is building a product that uses AI to a credit agency's AI scoring tool whose output is shared with third parties to decide on housing or telecom access.,high_risk | |
| Software evaluating coding-bootcamp learners' project submissions to determine their final certification outcome.,high_risk | |
| A startup is building a product that uses AI to an online banking app verifying a customer's identity via facial scan to let them log in.,minimal_risk | |
| A startup is building a product that uses AI to software scheduling campaign rallies and assigning volunteers to shifts for a political party.,minimal_risk | |
| "We're planning to use machine learning to an AI tool prioritising social-housing applicants by income, household size and urgency.",high_risk | |
| A smart speaker that identifies itself by name as an AI assistant when a user first sets it up.,limited_risk | |
| "Our client wants to deploy an AI tool to software verifying that a passport's chip and security features are genuine, unrelated to detecting people.",minimal_risk | |
| A startup is building a product that uses AI to a tool generating placeholder text for a website design mockup.,minimal_risk | |
| "A startup is building a product that uses AI to a tool helping a loan applicant understand and fill out a credit application form, separate from the scoring itself.",minimal_risk | |
| A startup is building a product that uses AI to a marketing tool de-duplicating a mailing list before a newsletter send.,minimal_risk | |
| "A startup is building a product that uses AI to software verifying that a passport's chip and security features are genuine, unrelated to detecting people.",minimal_risk | |
| A startup is building a product that uses AI to a car safety system detecting driver drowsiness or eye closure to trigger an alert and prevent accidents.,minimal_risk | |
| Proposed use case: an AI system designed to a university exam-monitoring tool that reads students' facial emotions during tests to flag 'suspicious' emotional states.,prohibited | |
| "A startup is building a product that uses AI to software ranking neighbourhood residents by predicted criminality using only inferred psychological traits, no factual record.",prohibited | |
| We are developing an AI system that will a code-completion tool suggesting the next line while a developer writes software.,minimal_risk | |
| We are developing an AI system that will an AI tool comparing footage from a private security camera against a database to identify a suspect seen on the recording.,high_risk | |
| A city government running live facial recognition across all town-centre CCTV cameras to identify any passer-by for general policing.,prohibited | |
| "A startup is building a product that uses AI to software checking that a visa application form has all required fields filled in, without assessing its content.",minimal_risk | |
| An insurer's AI pricing individual life-insurance premiums based on predicted health risk.,high_risk | |
| Software computing an individual security or health risk flag during visa processing for named applicants.,high_risk | |
| An AI assistant drafting a first pass of internal meeting minutes for review.,minimal_risk | |
| A startup is building a product that uses AI to a video-editing tool generating a synthetic video of a spokesperson delivering a scripted ad.,limited_risk | |
| A startup is building a product that uses AI to a university exam-monitoring tool that reads students' facial emotions during tests to flag 'suspicious' emotional states.,prohibited | |
| A startup is building a product that uses AI to a vending machine using biometric age estimation to decide whether to sell an age-restricted product.,high_risk | |
| Our client wants to deploy an AI tool to a personal photo app auto-tagging holiday pictures with generic labels like 'beach' or 'mountains'.,minimal_risk | |
| A startup is building a product that uses AI to a shopping mall's live facial recognition system feeding continuous passer-by identification to police with no judicial authorisation.,prohibited | |
| Proposed use case: an AI system designed to a startup scraping public social media photos at scale to build a facial recognition database without the people's knowledge or consent.,prohibited | |
| "Proposed use case: an AI system designed to an AI tool forecasting electricity demand for grid capacity planning, separate from any safety control system.",minimal_risk | |
| A startup is building a product that uses AI to a farm-management tool predicting optimal irrigation timing from soil-moisture sensors.,minimal_risk | |
| An immigration system assessing the security risk posed by a person applying to enter the country.,high_risk | |
| We are developing an AI system that will an AI assistant drafting a first pass of internal meeting minutes for review.,minimal_risk | |
| "A startup is building a product that uses AI to an AI tool prioritising social-housing applicants by income, household size and urgency.",high_risk | |
| "We are developing an AI system that will software verifying a job applicant's professional licence number against an official registry, output is just confirmed/not-confirmed.",minimal_risk | |
| Our client wants to deploy an AI tool to a remote-work monitoring tool reading employees' facial expressions during video calls to report 'engagement scores' to managers.,prohibited | |
| Proposed use case: an AI system designed to a system assigning daily task loads to warehouse workers based on automated behavioural profiling.,high_risk | |
| "Proposed use case: an AI system designed to software recommending which government office should review a benefits application, without assessing the application itself.",minimal_risk | |
| Proposed use case: an AI system designed to software predicting which specific person in a neighbourhood is most likely to commit burglary based solely on demographic profiling.,prohibited | |
| "We are developing an AI system that will software monitoring which bills an elected official has sponsored, for public transparency reporting.",minimal_risk | |
| Proposed use case: an AI system designed to an AI tool used to decide employee promotions from automated productivity scoring alone.,high_risk | |
| Proposed use case: an AI system designed to an AI-based smartphone unlock feature using the owner's own face to authenticate.,minimal_risk | |
| Software analysing online posts' sentiment and networks to flag users as potential extremist-affiliated persons.,high_risk | |
| "Proposed use case: an AI system designed to an outreach tool proactively identifying people who might benefit from preventive healthcare, without deciding eligibility.",minimal_risk | |
| Our client wants to deploy an AI tool to an online banking app verifying a customer's identity via facial scan to let them log in.,minimal_risk | |
| A startup is building a product that uses AI to software categorising conference attendees by inferred demographic traits based on facial biometric data.,high_risk | |
| We are developing an AI system that will software matching CCTV facial images from a stadium incident against a database to identify offenders after the fact.,high_risk | |
| A credit agency's AI scoring tool whose output is shared with third parties to decide on housing or telecom access.,high_risk | |
| An app that identifies users with cognitive disabilities and adjusts its interface to make them more likely to accept unfavourable terms.,prohibited | |
| A startup is building a product that uses AI to software used by prosecutors to automatically evaluate witness statement reliability in an active case.,high_risk | |
| "Software predicting which city blocks have higher burglary rates, without assessing risk to any named individual.",minimal_risk | |
| A startup is building a product that uses AI to software used by a public agency to decide whether to revoke someone's disability support payments.,high_risk | |
| We're planning to use machine learning to software reviewing photos of installed electricity meters and flagging installation errors for a technician to check.,minimal_risk | |
| A startup is building a product that uses AI to a chatbot trained to deceptively impersonate a trusted friend's writing style to talk someone into a financial decision against their interest.,prohibited | |
| "An airport's live facial recognition scanning every traveller in real time for general security monitoring, no judicial authorisation.",prohibited | |
| We are developing an AI system that will software estimating shipping costs based on package weight and distance.,minimal_risk | |
| A startup is building a product that uses AI to a parole-support tool combining verified criminal record data with a caseworker's assessment to estimate reoffending risk.,high_risk | |
| Proposed use case: an AI system designed to an airline's chatbot handling flight-change requests in direct conversation with passengers.,limited_risk | |
| "Proposed use case: an AI system designed to an automated tool flagging known child-sexual-abuse-material files online for removal, without assessing evidentiary reliability.",minimal_risk | |
| "Our client wants to deploy an AI tool to predictive-maintenance software flagging infrastructure components likely to need servicing soon, as one precaution among several.",minimal_risk | |
| A startup is building a product that uses AI to a court-assistance tool helping a judge research relevant case law and apply it to the facts of a pending case.,high_risk | |
| A school hallway camera system flagging bullying incidents outside of any testing context.,minimal_risk | |
| "A stadium entry-gate system checking a visitor's face against their own ticket to confirm identity, not to search a watchlist.",minimal_risk | |
| We're planning to use machine learning to software generating draft small-claims rulings from structured case data and standard legal templates.,high_risk | |
| We're planning to use machine learning to a tool generating placeholder text for a website design mockup.,minimal_risk | |
| A startup is building a product that uses AI to a customer-support chatbot on an e-commerce site answering questions directly with shoppers.,limited_risk | |
| Proposed use case: an AI system designed to software used by law enforcement to flag which persons are statistically likely to become crime victims.,high_risk | |
| Proposed use case: an AI system designed to a credit agency's AI scoring tool whose output is shared with third parties to decide on housing or telecom access.,high_risk | |
| We are developing an AI system that will a shift-scheduling AI that ranks warehouse workers by performance and allocates the best shifts accordingly.,high_risk | |
| Our client wants to deploy an AI tool to an AI system managing traffic-signal timing across a city's road network as a core safety function.,high_risk | |
| Our client wants to deploy an AI tool to an AI system deciding whether an applicant qualifies for unemployment benefits and in what amount.,high_risk | |
| A streaming service recommending shows based only on a user's own viewing history on that same service.,minimal_risk | |
| We're planning to use machine learning to a car safety system detecting driver drowsiness or eye closure to trigger an alert and prevent accidents.,minimal_risk | |
| A startup is building a product that uses AI to a voiceprint-matching system identifying speakers by comparing recordings against a database of known individuals.,high_risk | |
| "A company harvesting profile pictures from across the internet, untargeted, to train and grow a face-matching product.",prohibited | |
| Our client wants to deploy an AI tool to a landlord-screening platform combining tenants' social media posts and unrelated purchase history to deny housing applications.,prohibited | |
| We're planning to use machine learning to a banking app's AI assistant that answers account questions in a live chat with customers.,limited_risk | |
| "A startup is building a product that uses AI to a job-matching tool that scores and ranks candidates from CVs, generating a 'top 5' shortlist for recruiters.",high_risk | |
| We are developing an AI system that will an age-verification system estimating a user's age bracket from a selfie to restrict access to age-rated content.,high_risk | |
| Our client wants to deploy an AI tool to an AI system on a surveillance tower flagging human presence near a land border to trigger patrol dispatch.,high_risk | |
| Our client wants to deploy an AI tool to software optimising delivery-truck routes to reduce fuel use.,minimal_risk | |
| Proposed use case: an AI system designed to a 911/112 dispatch system that automatically prioritises which emergency calls get a first responder sent first.,high_risk | |
| We're planning to use machine learning to an AI system producing a per-traveller risk score from travel history and watch-list matches for border referral.,high_risk | |
| A chatbot trained to deceptively impersonate a trusted friend's writing style to talk someone into a financial decision against their interest.,prohibited | |
| Our client wants to deploy an AI tool to software flagging potential typos in a document before publishing.,minimal_risk | |
| Our client wants to deploy an AI tool to software allocating limited home-care hours among elderly applicants based on need.,high_risk | |
| An internal ticket-routing tool sorting IT support requests by category.,minimal_risk | |
| "Our client wants to deploy an AI tool to an ad-tech platform using standard persuasive copywriting techniques to encourage purchases, fully compliant with consumer law.",minimal_risk | |
| Our client wants to deploy an AI tool to an AI pressure sensor embedded in a drinking-water distribution monitoring system.,high_risk | |
| We're planning to use machine learning to a polygraph-style AI tool used during a police interrogation to assess truthfulness.,high_risk | |
| Proposed use case: an AI system designed to a system inferring shoppers' age bracket and likely income group from biometric scans for retail analytics.,high_risk | |
| A system assigning daily task loads to warehouse workers based on automated behavioural profiling.,high_risk | |
| A game with dark-pattern AI that manipulates a player's emotional state to maximise in-game purchases beyond what they intended to spend.,prohibited | |
| "We are developing an AI system that will a stadium entry-gate system checking a visitor's face against their own ticket to confirm identity, not to search a watchlist.",minimal_risk | |
| We are developing an AI system that will an app estimating a shopper's age bracket from a photo purely to personalise in-store advertising.,minimal_risk | |
| A call-taker assistant that listens for signs of a life-threatening emergency and helps classify call severity.,high_risk | |
| A startup is building a product that uses AI to a bank's AI system calculating a loan applicant's credit score to approve or reject a mortgage.,high_risk | |
| Our client wants to deploy an AI tool to a fintech app establishing a personal creditworthiness score used to set someone's borrowing limit.,high_risk | |
| "A startup is building a product that uses AI to software assisting a judge's sentencing decision using documented case facts alongside a risk score, not personality alone.",high_risk | |
| Our client wants to deploy an AI tool to software reviewing photos of installed electricity meters and flagging installation errors for a technician to check.,minimal_risk | |
| Our client wants to deploy an AI tool to a smart speaker that identifies itself by name as an AI assistant when a user first sets it up.,limited_risk | |
| We're planning to use machine learning to software used by prosecutors to automatically evaluate witness statement reliability in an active case.,high_risk | |
| We're planning to use machine learning to a court-assistance tool helping a judge research relevant case law and apply it to the facts of a pending case.,high_risk | |
| "A retailer running a legal targeted-advertising campaign based on browsing history, using ordinary marketing persuasion.",minimal_risk | |
| Proposed use case: an AI system designed to a government system automatically deciding eligibility for unemployment benefit applications.,high_risk | |
| Our client wants to deploy an AI tool to an age-verification system estimating a user's age bracket from a selfie to restrict access to age-rated content.,high_risk | |
| "We are developing an AI system that will a tool helping a loan applicant understand and fill out a credit application form, separate from the scoring itself.",minimal_risk | |
| Our client wants to deploy an AI tool to an AI tool comparing footage from a private security camera against a database to identify a suspect seen on the recording.,high_risk | |
| A startup is building a product that uses AI to a service that identifies unnamed individuals in old CCTV recordings by matching their faces to a reference database.,high_risk | |
| "We're planning to use machine learning to an airport's live facial recognition scanning every traveller in real time for general security monitoring, no judicial authorisation.",prohibited | |
| "Our client wants to deploy an AI tool to a call-centre tool detecting customer frustration from voice tone to route calls, outside any workplace/education context.",high_risk | |
| Our client wants to deploy an AI tool to a startup scraping public social media photos at scale to build a facial recognition database without the people's knowledge or consent.,prohibited | |
| We are developing an AI system that will software used by prosecutors to automatically evaluate witness statement reliability in an active case.,high_risk | |
| We're planning to use machine learning to software generating a synthetic voiceover that sounds like a real professional narrator for an audiobook.,limited_risk | |
| Proposed use case: an AI system designed to a voice assistant that uses manipulative conversational tactics designed to wear down an elderly user's resistance until they agree to a costly subscription.,prohibited | |
| Proposed use case: an AI system designed to a border agency's polygraph-style AI tool used during traveller interviews at a checkpoint.,high_risk | |
| "Proposed use case: an AI system designed to software checking that a visa application form has all required fields filled in, without assessing its content.",minimal_risk | |
| "A startup is building a product that uses AI to an app recommending university programmes to a student based on their own stated interests, with no role in admissions.",minimal_risk | |
| We're planning to use machine learning to a recruitment tool inferring candidates' religious beliefs from headshot photos to pass to hiring managers.,prohibited | |
| "Predictive-maintenance software flagging infrastructure components likely to need servicing soon, as one precaution among several.",minimal_risk | |
| We are developing an AI system that will an online store suggesting related products based on a shopper's own past purchases on that same store.,minimal_risk | |
| Proposed use case: an AI system designed to an insurer's AI pricing individual life-insurance premiums based on predicted health risk.,high_risk | |
| A VR headset app using imperceptible visual stimuli to steer a user's in-game purchase decisions without their awareness.,prohibited | |
| A startup is building a product that uses AI to an AI system on a surveillance tower flagging human presence near a land border to trigger patrol dispatch.,high_risk | |
| We're planning to use machine learning to an employer scoring staff using unrelated personal social behaviour to deny internal transfer requests disproportionately.,prohibited | |
| We are developing an AI system that will an immigration system assessing the security risk posed by a person applying to enter the country.,high_risk | |
| "A startup is building a product that uses AI to a pronunciation-practice tool giving students feedback on their accent, not used by teachers to grade them.",minimal_risk | |
| We're planning to use machine learning to software using flicker rates below conscious perception embedded in an ad to increase impulse buying.,prohibited | |
| "Proposed use case: an AI system designed to software predicting which city blocks have higher burglary rates, without assessing risk to any named individual.",minimal_risk | |
| "Our client wants to deploy an AI tool to a lecture-hall camera system inferring student boredom levels for a professor's dashboard, unrelated to any safety concern.",prohibited | |
| We're planning to use machine learning to a transit authority scanning every commuter's face in real time against a general watchlist for routine law enforcement purposes.,prohibited | |
| We're planning to use machine learning to law-enforcement software analysing physiological signals as a lie-detection aid during questioning.,high_risk | |
| We're planning to use machine learning to software used by law enforcement to flag which persons are statistically likely to become crime victims.,high_risk | |
| We are developing an AI system that will a fintech app establishing a personal creditworthiness score used to set someone's borrowing limit.,high_risk | |
| Proposed use case: an AI system designed to an AI system managing traffic-signal timing across a city's road network as a core safety function.,high_risk | |
| Software assigning primary-school children to specific classes or streams based on an automated assessment.,high_risk | |
| A startup is building a product that uses AI to a system that identifies elderly users showing signs of confusion and increases pressure tactics in a sales call script accordingly.,prohibited | |
| We're planning to use machine learning to software analysing online posts' sentiment and networks to flag users as potential extremist-affiliated persons.,high_risk | |
| Proposed use case: an AI system designed to a tool generating placeholder text for a website design mockup.,minimal_risk | |
| "Proposed use case: an AI system designed to a stadium entry-gate system checking a visitor's face against their own ticket to confirm identity, not to search a watchlist.",minimal_risk | |
| "A border-control AI system used to identify individuals during migration processing, beyond simple travel-document checks.",high_risk | |
| Proposed use case: an AI system designed to a game with dark-pattern AI that manipulates a player's emotional state to maximise in-game purchases beyond what they intended to spend.,prohibited | |
| Our client wants to deploy an AI tool to a service that identifies unnamed individuals in old CCTV recordings by matching their faces to a reference database.,high_risk | |
| "We are developing an AI system that will an automated tool flagging known child-sexual-abuse-material files online for removal, without assessing evidentiary reliability.",minimal_risk | |
| Proposed use case: an AI system designed to software predicting a risk group's expected annual medical cost to set the health-insurance premium.,high_risk | |
| Our client wants to deploy an AI tool to a VR headset app using imperceptible visual stimuli to steer a user's in-game purchase decisions without their awareness.,prohibited | |
| We're planning to use machine learning to an airline's chatbot handling flight-change requests in direct conversation with passengers.,limited_risk | |
| "A startup is building a product that uses AI to customer segmentation software analysing spending patterns for marketing, with no role in creditworthiness assessment.",minimal_risk | |
| "A website's help widget clearly labelled 'AI Assistant' at the top of the chat window, answering visitor questions.",limited_risk | |
| A startup is building a product that uses AI to an AI system managing traffic-signal timing across a city's road network as a core safety function.,high_risk | |
| Proposed use case: an AI system designed to a local news site auto-publishing AI-written articles about public interest topics with no editorial review.,limited_risk | |
| We are developing an AI system that will a system inferring shoppers' age bracket and likely income group from biometric scans for retail analytics.,high_risk | |
| "We are developing an AI system that will customer segmentation software analysing spending patterns for marketing, with no role in creditworthiness assessment.",minimal_risk | |
| We are developing an AI system that will a grammar checker suggesting edits in a word processor.,minimal_risk | |
| Software deciding which employees are selected for layoffs based on an automated performance model.,high_risk | |
| "Our client wants to deploy an AI tool to a website's help widget clearly labelled 'AI Assistant' at the top of the chat window, answering visitor questions.",limited_risk | |
| Proposed use case: an AI system designed to a system that detects members of a specific ethnic minority and steers them toward disadvantageous contract terms.,prohibited | |
| We're planning to use machine learning to an age-verification system estimating a user's age bracket from a selfie to restrict access to age-rated content.,high_risk | |
| Our client wants to deploy an AI tool to image-generation software producing photorealistic product renders used in an online catalogue.,limited_risk | |
| A startup is building a product that uses AI to a city government running live facial recognition across all town-centre CCTV cameras to identify any passer-by for general policing.,prohibited | |
| "We're planning to use machine learning to a physical rehabilitation app adjusting exercise difficulty based on patient progress, under a doctor's supervision.",minimal_risk | |
| "We are developing an AI system that will a system that flags a named individual as a likely future shoplifter based only on their personality profile, with no history of offences.",prohibited | |
| Our client wants to deploy an AI tool to a language-learning app giving instant feedback on exercises that don't count toward any certification.,minimal_risk | |
| A startup is building a product that uses AI to a startup scraping public social media photos at scale to build a facial recognition database without the people's knowledge or consent.,prohibited | |
| We're planning to use machine learning to a VR headset app using imperceptible visual stimuli to steer a user's in-game purchase decisions without their awareness.,prohibited | |
| "Proposed use case: an AI system designed to an AI tool prioritising social-housing applicants by income, household size and urgency.",high_risk | |
| We are developing an AI system that will software cross-referencing databases to generate a decision on an individual's eligibility for social benefits.,high_risk | |
| We are developing an AI system that will an airline's chatbot handling flight-change requests in direct conversation with passengers.,limited_risk | |
| Our client wants to deploy an AI tool to an employer scoring staff using unrelated personal social behaviour to deny internal transfer requests disproportionately.,prohibited | |
| We're planning to use machine learning to software cross-referencing databases to generate a decision on an individual's eligibility for social benefits.,high_risk | |
| We are developing an AI system that will an AI tool assessing the authenticity and consistency of an asylum applicant's documents and personal narrative.,high_risk | |
| A recruitment tool inferring candidates' religious beliefs from headshot photos to pass to hiring managers.,prohibited | |
| Our client wants to deploy an AI tool to an AI system estimating how many people are waiting at an airport gate to balance queue lanes.,minimal_risk | |
| Proposed use case: an AI system designed to software cross-referencing databases to generate a decision on an individual's eligibility for social benefits.,high_risk | |
| Proposed use case: an AI system designed to a spam filter sorting unwanted emails out of a company inbox.,minimal_risk | |
| We're planning to use machine learning to a voiceprint-matching system identifying speakers by comparing recordings against a database of known individuals.,high_risk | |
| We are developing an AI system that will a streaming service recommending shows based only on a user's own viewing history on that same service.,minimal_risk | |
| "We are developing an AI system that will a call-centre tool detecting customer frustration from voice tone to route calls, outside any workplace/education context.",high_risk | |
| "A startup is building a product that uses AI to an ad-tech platform using standard persuasive copywriting techniques to encourage purchases, fully compliant with consumer law.",minimal_risk | |
| Our client wants to deploy an AI tool to software supporting a probation officer's parole recommendation by evaluating criminal record and behaviour.,high_risk | |
| Our client wants to deploy an AI tool to a call-taker assistant that listens for signs of a life-threatening emergency and helps classify call severity.,high_risk | |
| We are developing an AI system that will software generating a synthetic voiceover that sounds like a real professional narrator for an audiobook.,limited_risk | |
| A startup is building a product that uses AI to a system categorising job applicants' trade union sympathies by analysing their facial biometric data from video interviews.,prohibited | |
| A marketing team using AI to generate product photos that look like real studio photography.,limited_risk | |
| "Our client wants to deploy an AI tool to a plagiarism checker comparing submitted homework against a database of existing text, done after submission.",minimal_risk | |
| We're planning to use machine learning to a school hallway camera system flagging bullying incidents outside of any testing context.,minimal_risk | |
| A customer-support chatbot on an e-commerce site answering questions directly with shoppers.,limited_risk | |
| An AI pressure sensor embedded in a drinking-water distribution monitoring system.,high_risk | |
| A startup is building a product that uses AI to a border agency's polygraph-style AI tool used during traveller interviews at a checkpoint.,high_risk | |
| "We're planning to use machine learning to software assisting a judge's sentencing decision using documented case facts alongside a risk score, not personality alone.",high_risk | |
| A startup is building a product that uses AI to an image-compression tool automatically resizing product photos before upload.,minimal_risk | |
| We are developing an AI system that will AI software assisting an arbitration panel in interpreting contract law for a dispute resolution.,high_risk | |
| A startup is building a product that uses AI to an admissions algorithm deciding which applicants get into a university's engineering programme.,high_risk | |
| We're planning to use machine learning to a landlord-screening platform combining tenants' social media posts and unrelated purchase history to deny housing applications.,prohibited | |
| Our client wants to deploy an AI tool to a retail analytics tool inferring shoppers' political leanings from facial biometric scans in-store.,prohibited | |
| "We're planning to use machine learning to predictive-maintenance software flagging infrastructure components likely to need servicing soon, as one precaution among several.",minimal_risk | |
| "A startup is building a product that uses AI to a reoffending-risk model used by a parole board to help decide early release, based on more than just profiling.",high_risk | |
| Proposed use case: an AI system designed to software generating draft small-claims rulings from structured case data and standard legal templates.,high_risk | |
| A startup is building a product that uses AI to AI-based digital forensics software detecting whether CCTV footage submitted as evidence has been altered.,high_risk | |
| We're planning to use machine learning to a spam filter sorting unwanted emails out of a company inbox.,minimal_risk | |
| Proposed use case: an AI system designed to law-enforcement software analysing physiological signals as a lie-detection aid during questioning.,high_risk | |
| Proposed use case: an AI system designed to law-enforcement software profiling individuals during the detection phase of an ongoing criminal case.,high_risk | |
| A startup is building a product that uses AI to an AI safety component managing gas-supply pressure regulation for a district heating network.,high_risk | |
| "Our client wants to deploy an AI tool to a physical rehabilitation app adjusting exercise difficulty based on patient progress, under a doctor's supervision.",minimal_risk | |
| We're planning to use machine learning to a university chatbot answering prospective students' questions about admission requirements and deadlines.,minimal_risk | |
| A startup is building a product that uses AI to an asylum agency's AI tool assisting officials in examining the eligibility of asylum applications.,high_risk | |
| Proposed use case: an AI system designed to software computing an individual security or health risk flag during visa processing for named applicants.,high_risk | |
| "Our client wants to deploy an AI tool to an app recommending university programmes to a student based on their own stated interests, with no role in admissions.",minimal_risk | |
| Software predicting a risk group's expected annual medical cost to set the health-insurance premium.,high_risk | |
| "We are developing an AI system that will an AI tool organising incoming CVs into a searchable internal database for recruiters, without scoring or ranking them.",minimal_risk | |
| A startup is building a product that uses AI to a control system for a municipal water treatment plant's safety-critical valves and monitoring.,high_risk | |
| Our client wants to deploy an AI tool to software predicting which specific person in a neighbourhood is most likely to commit burglary based solely on demographic profiling.,prohibited | |
| An automated essay-grading system that decides whether a student passes or fails a course.,high_risk | |
| "Our client wants to deploy an AI tool to an AI tool organising incoming CVs into a searchable internal database for recruiters, without scoring or ranking them.",minimal_risk | |
| We are developing an AI system that will software categorising conference attendees by inferred demographic traits based on facial biometric data.,high_risk | |
| "We're planning to use machine learning to a website's help widget clearly labelled 'AI Assistant' at the top of the chat window, answering visitor questions.",limited_risk | |
| A startup is building a product that uses AI to a voter-advice app recommending which party or candidate best matches a person's stated political views.,high_risk | |
| We're planning to use machine learning to an AI safety component managing gas-supply pressure regulation for a district heating network.,high_risk | |
| We're planning to use machine learning to a chatbot trained to deceptively impersonate a trusted friend's writing style to talk someone into a financial decision against their interest.,prohibited | |
| Our client wants to deploy an AI tool to a system categorising job applicants' trade union sympathies by analysing their facial biometric data from video interviews.,prohibited | |
| A startup is building a product that uses AI to a tool assessing the credibility of witness statements by checking internal consistency against verified facts.,high_risk | |
| Proposed use case: an AI system designed to image-generation software producing photorealistic product renders used in an online catalogue.,limited_risk | |
| We are developing an AI system that will a cloud provider's load-balancer distributing traffic across internal servers.,minimal_risk | |
| We're planning to use machine learning to software deciding which employees are selected for layoffs based on an automated performance model.,high_risk | |
| Proposed use case: an AI system designed to a police-used tool predicting which specific individuals are at risk of becoming victims of trafficking.,high_risk | |
| "We are developing an AI system that will software ranking neighbourhood residents by predicted criminality using only inferred psychological traits, no factual record.",prohibited | |
| A grammar checker suggesting edits in a word processor.,minimal_risk | |
| We are developing an AI system that will maritime AI tracking and detecting persons for border-control interception operations.,high_risk | |
| "A startup is building a product that uses AI to a lecture-hall camera system inferring student boredom levels for a professor's dashboard, unrelated to any safety concern.",prohibited | |
| We are developing an AI system that will software allocating limited home-care hours among elderly applicants based on need.,high_risk | |
| "We are developing an AI system that will software checking that a visa application form has all required fields filled in, without assessing its content.",minimal_risk | |
| Our client wants to deploy an AI tool to law-enforcement software analysing physiological signals as a lie-detection aid during questioning.,high_risk | |
| "Proposed use case: an AI system designed to an ad-tech platform using standard persuasive copywriting techniques to encourage purchases, fully compliant with consumer law.",minimal_risk | |
| "We're planning to use machine learning to an app recommending university programmes to a student based on their own stated interests, with no role in admissions.",minimal_risk | |
| Software assisting a visa office in assessing the reliability of evidence submitted with a residence-permit application.,high_risk | |
| Our client wants to deploy an AI tool to predictive maintenance software forecasting when a factory machine needs servicing from vibration sensors.,minimal_risk | |
| Software estimating shipping costs based on package weight and distance.,minimal_risk | |
| Proposed use case: an AI system designed to a personal photo app auto-tagging holiday pictures with generic labels like 'beach' or 'mountains'.,minimal_risk | |
| Our client wants to deploy an AI tool to a recruitment tool inferring candidates' religious beliefs from headshot photos to pass to hiring managers.,prohibited | |
| We are developing an AI system that will a control system for a municipal water treatment plant's safety-critical valves and monitoring.,high_risk | |
| We are developing an AI system that will a banking app's AI assistant that answers account questions in a live chat with customers.,limited_risk | |
| An AI system deciding whether an applicant qualifies for unemployment benefits and in what amount.,high_risk | |
| We're planning to use machine learning to a voice assistant that uses manipulative conversational tactics designed to wear down an elderly user's resistance until they agree to a costly subscription.,prohibited | |
| Our client wants to deploy an AI tool to an AI tool assessing the authenticity and consistency of an asylum applicant's documents and personal narrative.,high_risk | |
| "A pronunciation-practice tool giving students feedback on their accent, not used by teachers to grade them.",minimal_risk | |
| "We are developing an AI system that will a marketing team A/B testing ad headlines to see which one converts better, using normal legitimate techniques.",minimal_risk | |
| We're planning to use machine learning to a system categorising job applicants' trade union sympathies by analysing their facial biometric data from video interviews.,prohibited | |
| Our client wants to deploy an AI tool to software matching CCTV facial images from a stadium incident against a database to identify offenders after the fact.,high_risk | |
| Our client wants to deploy an AI tool to an AI tool allocating law-firm case work to associates based on a score built from billing hours and responsiveness.,high_risk | |
| Our client wants to deploy an AI tool to an image-compression tool automatically resizing product photos before upload.,minimal_risk | |
| An AI tool used to decide employee promotions from automated productivity scoring alone.,high_risk | |
| "Proposed use case: an AI system designed to a system that flags a named individual as a likely future shoplifter based only on their personality profile, with no history of offences.",prohibited | |
| "Proposed use case: an AI system designed to a tool assigning a 'future offender' risk label to a named person based purely on inferred personality traits, with no verifiable facts.",prohibited | |
| Proposed use case: an AI system designed to a streaming service recommending shows based only on a user's own viewing history on that same service.,minimal_risk | |
| An AI-based smartphone unlock feature using the owner's own face to authenticate.,minimal_risk | |
| A transit authority scanning every commuter's face in real time against a general watchlist for routine law enforcement purposes.,prohibited | |
| We're planning to use machine learning to a farm-management tool predicting optimal irrigation timing from soil-moisture sensors.,minimal_risk | |
| A parole-support tool combining verified criminal record data with a caseworker's assessment to estimate reoffending risk.,high_risk | |
| A startup is building a product that uses AI to an app that scans a person's face to infer their likely religious affiliation for targeted advertising.,prohibited | |
| A shopping mall's live facial recognition system feeding continuous passer-by identification to police with no judicial authorisation.,prohibited | |
| Our client wants to deploy an AI tool to an AI system producing a numerical creditworthiness score used to decide on a consumer loan or mortgage.,high_risk | |
| Proposed use case: an AI system designed to an AI tool ranking self-employed service providers on a platform to decide who gets shown to consumers.,high_risk | |
| Our client wants to deploy an AI tool to a vending machine using biometric age estimation to decide whether to sell an age-restricted product.,high_risk | |
| A startup is building a product that uses AI to an AI tool matching apprenticeship applicants to available vocational training slots based on qualifications.,high_risk | |
| A startup is building a product that uses AI to a private security firm's tool labelling named individuals as 'high crime risk' from personality-test results alone.,prohibited | |
| "We're planning to use machine learning to an automated tool flagging known child-sexual-abuse-material files online for removal, without assessing evidentiary reliability.",minimal_risk | |
| "We're planning to use machine learning to a retail store using AI to infer shoppers' mood from in-store cameras, disclosed via a sign at the entrance.",limited_risk | |
| "A marketing team A/B testing ad headlines to see which one converts better, using normal legitimate techniques.",minimal_risk | |
| "A startup is building a product that uses AI to a border-control AI system used to identify individuals during migration processing, beyond simple travel-document checks.",high_risk | |
| A university chatbot answering prospective students' questions about admission requirements and deadlines.,minimal_risk | |
| "We're planning to use machine learning to a company harvesting profile pictures from across the internet, untargeted, to train and grow a face-matching product.",prohibited | |
| Proposed use case: an AI system designed to a city government running live facial recognition across all town-centre CCTV cameras to identify any passer-by for general policing.,prohibited | |
| "Our client wants to deploy an AI tool to software flagging unusual transaction patterns that may indicate money laundering, based only on amounts and timing, not personal profiles.",minimal_risk | |
| Our client wants to deploy an AI tool to a border agency's polygraph-style AI tool used during traveller interviews at a checkpoint.,high_risk | |
| "We are developing an AI system that will software predicting which city blocks have higher burglary rates, without assessing risk to any named individual.",minimal_risk | |
| We're planning to use machine learning to an automated tool determining eligibility for subsidised healthcare enrollment.,high_risk | |
| "A startup is building a product that uses AI to software monitoring which bills an elected official has sponsored, for public transparency reporting.",minimal_risk | |
| "Software flagging unusual transaction patterns that may indicate money laundering, based only on amounts and timing, not personal profiles.",minimal_risk | |
| Proposed use case: an AI system designed to an app that identifies users with cognitive disabilities and adjusts its interface to make them more likely to accept unfavourable terms.,prohibited | |
| We are developing an AI system that will law-enforcement software profiling individuals during the detection phase of an ongoing criminal case.,high_risk | |
| A startup is building a product that uses AI to an app that uses flashing subliminal visual cues below conscious perception to push users toward impulse purchases they later regret.,prohibited | |
| "We are developing an AI system that will a reoffending-risk model used by a parole board to help decide early release, based on more than just profiling.",high_risk | |
| "A startup is building a product that uses AI to a plagiarism checker comparing submitted homework against a database of existing text, done after submission.",minimal_risk | |
| "We are developing an AI system that will software predicting which neighbourhood and time window has elevated burglary risk, without linking to a specific person.",minimal_risk | |
| Our client wants to deploy an AI tool to matching a person's face after the fact against a security-camera watchlist to identify them from footage.,high_risk | |
| Proposed use case: an AI system designed to an AI system estimating how many people are waiting at an airport gate to balance queue lanes.,minimal_risk | |
| Our client wants to deploy an AI tool to a control system for a municipal water treatment plant's safety-critical valves and monitoring.,high_risk | |
| An AI tool allocating law-firm case work to associates based on a score built from billing hours and responsiveness.,high_risk | |
| "A retail store using AI to infer shoppers' mood from in-store cameras, disclosed via a sign at the entrance.",limited_risk | |
| Software supporting a probation officer's parole recommendation by evaluating criminal record and behaviour.,high_risk | |
| Proposed use case: an AI system designed to a retail analytics tool inferring shoppers' political leanings from facial biometric scans in-store.,prohibited | |
| We are developing an AI system that will a university chatbot answering prospective students' questions about admission requirements and deadlines.,minimal_risk | |
| Our client wants to deploy an AI tool to a system assigning daily task loads to warehouse workers based on automated behavioural profiling.,high_risk | |
| "We are developing an AI system that will an insurer's AI reviewing life-insurance applications using age, health status and lifestyle to decide terms.",high_risk | |
| A startup is building a product that uses AI to maritime AI tracking and detecting persons for border-control interception operations.,high_risk | |
| A startup is building a product that uses AI to an AI-based smartphone unlock feature using the owner's own face to authenticate.,minimal_risk | |
| "We're planning to use machine learning to software predicting which neighbourhood and time window has elevated burglary risk, without linking to a specific person.",minimal_risk | |
| We are developing an AI system that will a remote-work monitoring tool reading employees' facial expressions during video calls to report 'engagement scores' to managers.,prohibited | |
| Our client wants to deploy an AI tool to an asylum agency's AI tool assisting officials in examining the eligibility of asylum applications.,high_risk | |
| We are developing an AI system that will software computing an individual security or health risk flag during visa processing for named applicants.,high_risk | |
| "An office system that reads employees' facial expressions to adjust meeting-room lighting, disclosed via posted notice.",limited_risk | |
| We're planning to use machine learning to software evaluating coding-bootcamp learners' project submissions to determine their final certification outcome.,high_risk | |
| "A startup is building a product that uses AI to an outreach tool proactively identifying people who might benefit from preventive healthcare, without deciding eligibility.",minimal_risk | |
| We are developing an AI system that will an AI learning assistant whose end-of-unit reports on student progress are used by teachers to set final grades.,high_risk | |
| Proposed use case: an AI system designed to a marketing team using AI to generate product photos that look like real studio photography.,limited_risk | |
| A smart thermostat learning a household's temperature preferences over time.,minimal_risk | |
| Proposed use case: an AI system designed to a transit authority scanning every commuter's face in real time against a general watchlist for routine law enforcement purposes.,prohibited | |
| We are developing an AI system that will software optimising delivery-truck routes to reduce fuel use.,minimal_risk | |
| Our client wants to deploy an AI tool to a private security firm's tool labelling named individuals as 'high crime risk' from personality-test results alone.,prohibited | |
| We are developing an AI system that will a polygraph-style AI tool used during a police interrogation to assess truthfulness.,high_risk | |
| "Proposed use case: an AI system designed to a retail store using AI to infer shoppers' mood from in-store cameras, disclosed via a sign at the entrance.",limited_risk | |
| We are developing an AI system that will an app that uses flashing subliminal visual cues below conscious perception to push users toward impulse purchases they later regret.,prohibited | |
| We are developing an AI system that will an AI system classifying 112 emergency calls by urgency and routing responders using natural language processing.,high_risk | |
| "A startup is building a product that uses AI to an office system that reads employees' facial expressions to adjust meeting-room lighting, disclosed via posted notice.",limited_risk | |
| We're planning to use machine learning to a language-learning app giving instant feedback on exercises that don't count toward any certification.,minimal_risk | |
| "Our client wants to deploy an AI tool to software verifying a job applicant's professional licence number against an official registry, output is just confirmed/not-confirmed.",minimal_risk | |
| We're planning to use machine learning to software optimising delivery-truck routes to reduce fuel use.,minimal_risk | |
| A startup is building a product that uses AI to matching a person's face after the fact against a security-camera watchlist to identify them from footage.,high_risk | |
| We are developing an AI system that will a university exam-monitoring tool that reads students' facial emotions during tests to flag 'suspicious' emotional states.,prohibited | |
| A fintech app establishing a personal creditworthiness score used to set someone's borrowing limit.,high_risk | |
| Proposed use case: an AI system designed to an AI system classifying 112 emergency calls by urgency and routing responders using natural language processing.,high_risk | |
| Our client wants to deploy an AI tool to a smart thermostat learning a household's temperature preferences over time.,minimal_risk | |
| A startup is building a product that uses AI to an AI assistant drafting a first pass of internal meeting minutes for review.,minimal_risk | |
| Proposed use case: an AI system designed to a toy that uses AI to exploit small children's trust to pressure them into asking parents to buy in-app add-ons.,prohibited | |
| Proposed use case: an AI system designed to an online store suggesting related products based on a shopper's own past purchases on that same store.,minimal_risk | |
| We are developing an AI system that will an employer scoring staff using unrelated personal social behaviour to deny internal transfer requests disproportionately.,prohibited | |
| Proposed use case: an AI system designed to an AI tool assessing the authenticity and consistency of an asylum applicant's documents and personal narrative.,high_risk | |
| "A startup is building a product that uses AI to an AI tool forecasting electricity demand for grid capacity planning, separate from any safety control system.",minimal_risk | |
| "A lecture-hall camera system inferring student boredom levels for a professor's dashboard, unrelated to any safety concern.",prohibited | |
| A marketing tool de-duplicating a mailing list before a newsletter send.,minimal_risk | |
| Proposed use case: an AI system designed to a smart speaker that identifies itself by name as an AI assistant when a user first sets it up.,limited_risk | |
| We're planning to use machine learning to matching a person's face after the fact against a security-camera watchlist to identify them from footage.,high_risk | |
| Proposed use case: an AI system designed to a language-learning app giving instant feedback on exercises that don't count toward any certification.,minimal_risk | |
| Our client wants to deploy an AI tool to an app estimating a shopper's age bracket from a photo purely to personalise in-store advertising.,minimal_risk | |
| A startup is building a product that uses AI to a brain-computer interface product that uses subliminal audio cues to push users toward a specific financial choice.,prohibited | |
| A startup is building a product that uses AI to a code-completion tool suggesting the next line while a developer writes software.,minimal_risk | |
| "Proposed use case: an AI system designed to a marketing team A/B testing ad headlines to see which one converts better, using normal legitimate techniques.",minimal_risk | |
| Proposed use case: an AI system designed to a video-editing tool generating a synthetic video of a spokesperson delivering a scripted ad.,limited_risk | |
| We are developing an AI system that will an AI system estimating how many people are waiting at an airport gate to balance queue lanes.,minimal_risk | |
| Live facial recognition at a border checkpoint comparing travellers against biometric watchlists in real time.,high_risk | |
| We're planning to use machine learning to a smart thermostat learning a household's temperature preferences over time.,minimal_risk | |
| Our client wants to deploy an AI tool to software used by law enforcement to flag which persons are statistically likely to become crime victims.,high_risk | |
| We're planning to use machine learning to an immigration system assessing the security risk posed by a person applying to enter the country.,high_risk | |
| "Proposed use case: an AI system designed to software ranking neighbourhood residents by predicted criminality using only inferred psychological traits, no factual record.",prohibited | |
| Our client wants to deploy an AI tool to an app that scans a person's face to infer their likely religious affiliation for targeted advertising.,prohibited | |
| We are developing an AI system that will a government system automatically deciding eligibility for unemployment benefit applications.,high_risk | |
| We are developing an AI system that will software supporting a probation officer's parole recommendation by evaluating criminal record and behaviour.,high_risk | |
| Our client wants to deploy an AI tool to a data broker continuously scraping millions of public Instagram photos to expand a commercial facial-recognition dataset.,prohibited | |
| Proposed use case: an AI system designed to software assisting a visa office in assessing the reliability of evidence submitted with a residence-permit application.,high_risk | |
| We're planning to use machine learning to a parole-support tool combining verified criminal record data with a caseworker's assessment to estimate reoffending risk.,high_risk | |
| Our client wants to deploy an AI tool to a 911/112 dispatch system that automatically prioritises which emergency calls get a first responder sent first.,high_risk | |
| Proposed use case: an AI system designed to a warehouse system forecasting next month's demand for a product line.,minimal_risk | |
| Proposed use case: an AI system designed to a bank's AI system calculating a loan applicant's credit score to approve or reject a mortgage.,high_risk | |
| A startup is building a product that uses AI to a banking app's AI assistant that answers account questions in a live chat with customers.,limited_risk | |
| "A system that flags a named individual as a likely future shoplifter based only on their personality profile, with no history of offences.",prohibited | |
| Our client wants to deploy an AI tool to a brain-computer interface product that uses subliminal audio cues to push users toward a specific financial choice.,prohibited | |
| "We are developing an AI system that will an office system that reads employees' facial expressions to adjust meeting-room lighting, disclosed via posted notice.",limited_risk | |
| A startup is building a product that uses AI to an AI tool enhancing a blurry CCTV image so a human investigator can examine it more clearly.,minimal_risk | |
| Proposed use case: an AI system designed to an AI tool enhancing a blurry CCTV image so a human investigator can examine it more clearly.,minimal_risk | |
| Our client wants to deploy an AI tool to an AI system producing a per-traveller risk score from travel history and watch-list matches for border referral.,high_risk | |
| We're planning to use machine learning to live facial recognition at a border checkpoint comparing travellers against biometric watchlists in real time.,high_risk | |
| Proposed use case: an AI system designed to a car safety system detecting driver drowsiness or eye closure to trigger an alert and prevent accidents.,minimal_risk | |
| "Our client wants to deploy an AI tool to an airport's live facial recognition scanning every traveller in real time for general security monitoring, no judicial authorisation.",prohibited | |
| "A startup is building a product that uses AI to an AI tool organising incoming CVs into a searchable internal database for recruiters, without scoring or ranking them.",minimal_risk | |
| We are developing an AI system that will a voiceprint-matching system identifying speakers by comparing recordings against a database of known individuals.,high_risk | |
| We are developing an AI system that will an image-compression tool automatically resizing product photos before upload.,minimal_risk | |
| Our client wants to deploy an AI tool to a game with dark-pattern AI that manipulates a player's emotional state to maximise in-game purchases beyond what they intended to spend.,prohibited | |
| A startup is building a product that uses AI to an AI tool allocating law-firm case work to associates based on a score built from billing hours and responsiveness.,high_risk | |
| A startup is building a product that uses AI to an AI pressure sensor embedded in a drinking-water distribution monitoring system.,high_risk | |
| "Software verifying a job applicant's professional licence number against an official registry, output is just confirmed/not-confirmed.",minimal_risk | |
| A startup is building a product that uses AI to live facial recognition at a border checkpoint comparing travellers against biometric watchlists in real time.,high_risk | |
| Our client wants to deploy an AI tool to a voter-advice app recommending which party or candidate best matches a person's stated political views.,high_risk | |
| A remote-work monitoring tool reading employees' facial expressions during video calls to report 'engagement scores' to managers.,prohibited | |
| Software predicting which specific person in a neighbourhood is most likely to commit burglary based solely on demographic profiling.,prohibited | |
| We're planning to use machine learning to software grading student exams and quizzes that count toward a final course evaluation.,high_risk | |
| We are developing an AI system that will software assigning primary-school children to specific classes or streams based on an automated assessment.,high_risk | |
| "We are developing an AI system that will software recommending which government office should review a benefits application, without assessing the application itself.",minimal_risk | |
| A data broker continuously scraping millions of public Instagram photos to expand a commercial facial-recognition dataset.,prohibited | |
| We're planning to use machine learning to an AI tool ranking self-employed service providers on a platform to decide who gets shown to consumers.,high_risk | |
| "We are developing an AI system that will a physical rehabilitation app adjusting exercise difficulty based on patient progress, under a doctor's supervision.",minimal_risk | |
| Our client wants to deploy an AI tool to AI-based digital forensics software detecting whether CCTV footage submitted as evidence has been altered.,high_risk | |
| We are developing an AI system that will a personal photo app auto-tagging holiday pictures with generic labels like 'beach' or 'mountains'.,minimal_risk | |
| A startup is building a product that uses AI to software reviewing photos of installed electricity meters and flagging installation errors for a technician to check.,minimal_risk | |
| A code-completion tool suggesting the next line while a developer writes software.,minimal_risk | |
| "Software verifying that a passport's chip and security features are genuine, unrelated to detecting people.",minimal_risk | |
| Proposed use case: an AI system designed to software used by a public agency to decide whether to revoke someone's disability support payments.,high_risk | |
| A landlord-screening platform combining tenants' social media posts and unrelated purchase history to deny housing applications.,prohibited | |
| We're planning to use machine learning to a system that identifies elderly users showing signs of confusion and increases pressure tactics in a sales call script accordingly.,prohibited | |
| We're planning to use machine learning to a data broker continuously scraping millions of public Instagram photos to expand a commercial facial-recognition dataset.,prohibited | |
| We're planning to use machine learning to AI software assisting an arbitration panel in interpreting contract law for a dispute resolution.,high_risk | |
| We are developing an AI system that will a video-editing tool generating a synthetic video of a spokesperson delivering a scripted ad.,limited_risk | |
| Proposed use case: an AI system designed to a system that identifies elderly users showing signs of confusion and increases pressure tactics in a sales call script accordingly.,prohibited | |
| A startup is building a product that uses AI to predictive maintenance software forecasting when a factory machine needs servicing from vibration sensors.,minimal_risk | |
| We are developing an AI system that will a customer-support chatbot on an e-commerce site answering questions directly with shoppers.,limited_risk | |
| Proposed use case: an AI system designed to software grading student exams and quizzes that count toward a final course evaluation.,high_risk | |
| A system inferring shoppers' age bracket and likely income group from biometric scans for retail analytics.,high_risk | |
| Proposed use case: an AI system designed to an AI system producing a numerical creditworthiness score used to decide on a consumer loan or mortgage.,high_risk | |
| Proposed use case: an AI system designed to an automated system determining scholarship eligibility based on an applicant's financial situation.,high_risk | |
| Our client wants to deploy an AI tool to an AI safety component managing gas-supply pressure regulation for a district heating network.,high_risk | |
| An AI system recognising heavy objects on a vulnerable bridge to prevent structural collapse.,high_risk | |
| "A startup is building a product that uses AI to a tool assigning a 'future offender' risk label to a named person based purely on inferred personality traits, with no verifiable facts.",prohibited | |
| We are developing an AI system that will an automated essay-grading system that decides whether a student passes or fails a course.,high_risk | |
| A voter-advice app recommending which party or candidate best matches a person's stated political views.,high_risk | |
| A startup is building a product that uses AI to an automated tool determining eligibility for subsidised healthcare enrollment.,high_risk | |
| "Proposed use case: an AI system designed to customer segmentation software analysing spending patterns for marketing, with no role in creditworthiness assessment.",minimal_risk | |
| Our client wants to deploy an AI tool to a grammar checker suggesting edits in a word processor.,minimal_risk | |
| Our client wants to deploy an AI tool to software evaluating coding-bootcamp learners' project submissions to determine their final certification outcome.,high_risk | |
| Our client wants to deploy an AI tool to an app that uses flashing subliminal visual cues below conscious perception to push users toward impulse purchases they later regret.,prohibited | |
| A startup is building a product that uses AI to software analysing online posts' sentiment and networks to flag users as potential extremist-affiliated persons.,high_risk | |
| We are developing an AI system that will an admissions algorithm deciding which applicants get into a university's engineering programme.,high_risk | |
| An AI tool ranking self-employed service providers on a platform to decide who gets shown to consumers.,high_risk | |
| A court-assistance tool helping a judge research relevant case law and apply it to the facts of a pending case.,high_risk | |
| "Our client wants to deploy an AI tool to software recommending which government office should review a benefits application, without assessing the application itself.",minimal_risk | |
| We are developing an AI system that will an AI system recognising heavy objects on a vulnerable bridge to prevent structural collapse.,high_risk | |
| We are developing an AI system that will software deciding which employees are selected for layoffs based on an automated performance model.,high_risk | |
| We are developing an AI system that will an AI system producing a numerical creditworthiness score used to decide on a consumer loan or mortgage.,high_risk | |
| Proposed use case: an AI system designed to a cloud provider's load-balancer distributing traffic across internal servers.,minimal_risk | |