Datasets:

Quome
/

consentbench

license: apache-2.0
pretty_name: ConsentBench
language:
  - en
tags:
  - clinical-ai
  - llm-security
  - ai-agents
  - healthcare
  - haarf
  - quokkaguard
  - synthetic
size_categories:
  - 1K<n<10K
task_categories:
  - other

ConsentBench

A FHIR-consent scope gate, AI-disclosure stamper, and signed competency gate for clinical AI agents.

ConsentBench is the benchmark corpus for Consent Is a Gateway Responsibility: Enforcing Patient Consent, AI Disclosure, and Participatory Governance for Clinical AI Agents, the quconsent module of the QUOKKAGUARD program. It ships with the quconsent repository, which contains the qfire gateway layer under test, the experiment harness, and the paper.

Four deterministic JSONL corpora under consentbench/: 200 FHIR-Consent-shaped directives, 712 scope/competency scenarios (in-scope, out-of-scope, explicit-deny, withdrawn, expired, draft; missing/uncredentialed/forged operators) with expected gateway behavior, 120 patient-facing prompts tagged with consent status, and 300 scored goals-of-care cases with two synthetic clinician raters; seed 42, no PHI.

All data are synthetic. No real patient data or protected health information (PHI) is included; clinical content is generated from templates with fixed seeds.

Files

File	Size	Rows
`consentbench/consent_directives.jsonl`	42 KB	200 rows
`consentbench/goals_of_care.jsonl`	50 KB	300 rows
`consentbench/patient_facing.jsonl`	29 KB	120 rows
`consentbench/scenarios.jsonl`	121 KB	712 rows

Record schemas

consentbench/consent_directives.jsonl: base, exceptions, patient, status
consentbench/goals_of_care.jsonl: action, case_id, label, patient, purpose, score, values
consentbench/patient_facing.jsonl: action, case_id, consent_status, expected, patient, prompt, purpose
consentbench/scenarios.jsonl: action, case_id, expected, family, operator, patient, purpose

How it was generated

The corpus is produced by the generator in the paper repository and is fully deterministic (fixed seeds), so it can be regenerated byte-for-byte.

# from the quconsent repository root (deterministic seeds)
python3 scripts/013-consent/gen_consent.py

Intended use

Evaluating the consent enforcement layer of a clinical-agent security gateway (HAARF control C4): An agent action can be authorized and clinically correct and still be unlawful: taken outside the patient's recorded consent, or delivered as a patient-facing message that hides its AI authorship.

The experiments that consume it (E-series in the paper) are reproduced from the repository:

git clone https://github.com/quome-cloud/quconsent
cd quconsent
cargo build --release

then follow the Reproduce the experiments section of its README.

Citation

This benchmark was built to evaluate a control of the Healthcare AI Agents Regulatory Framework (HAARF), the source framework for the QUOKKAGUARD program. Please cite both the paper and HAARF:

@unpublished{schwoebel2026quconsent,
  author = {{QUOKKAGUARD Working Group}},
  title  = {Consent Is a Gateway Responsibility: Enforcing Patient Consent, AI Disclosure, and Participatory Governance for Clinical AI Agents},
  note   = {Preprint. Quome, QUOKKAGUARD program (quconsent module)},
  year   = {2026},
  url    = {https://github.com/quome-cloud/quconsent}
}

@unpublished{schwoebel2026haarf,
  author = {Schwoebel, Jim and Frasch, Martin and Spalding, Art and Sewell, Ed and Englert, Phil and Halpert, Ben and Overbay, Collin and Semenec, Ingrida and Shor, Joel},
  title  = {{HAARF}: Healthcare {AI} agents regulatory framework --- a comprehensive security verification standard for autonomous {AI} systems in clinical environments},
  note   = {medRxiv Preprint},
  year   = {2026},
  month  = {April},
  doi    = {10.64898/2026.04.09.26350519},
  url    = {https://www.medrxiv.org/content/10.64898/2026.04.09.26350519v1}
}