Skip to content
Apart Research
Apart Research
Contact us
Sprint projectFeb 2, 2026London

Automated Compliance Measurement for Frontier AI Models: Evidence-Based Scoring of Model Card Disclosures

Yulong Lin · Team AI Transparency

Submitted to The Technical AI Governance Challenge. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Automated Compliance Measurement for Frontier AI Models: Evidence-Based Scoring of Model Card Disclosures

Open in a new tab (opens in new tab)
Code (opens in new tab)
Share

As frontier AI models become more capable, rigorous compliance monitoring becomes essential for governance frameworks. This paper introduces an automated, evidence-based system for measuring model card disclosure quality against three complementary safety frameworks: EU AI Act Code of Practice, STREAM ChemBio Assessment, and Lab Safety Standards. Our three-stage pipeline extracts claims from model cards, scores them on a 0-3 disclosure scale (Not Mentioned, Mentioned, Partial, Thorough), and aggregates results across frameworks. Validation against human expert annotation achieves perfect agreement (Cohen's κ = 1.0). Analyzing five frontier models reveals a consistent biosafety disclosure gap: average STREAM scores (59.8%) lag EU CoP scores (64.3%) by 4.6 percentage points across all models. Claude Opus 4.5 leads (69.6%), while disclosure quality varies substantially (range: 15.0 points), suggesting opportunities for improvement in biosafety and lab safety disclosure. Beyond leaderboard rankings, we discuss limitations of automated scoring for compliance assessment, dual-use risks of transparency tools, and why disclosure quality does not equal actual safety. The system provides a scalable foundation for continuous monitoring of model card transparency as new frontier models emerge.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for AI safety if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical AI safety problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.

  1. I hope i'm not misunderstanding this--- sorry, I think focusing so much on model cards is a huge map-territory problem. I don't know why I should trust that model cards are calibrated to or aligned with the models they describe, i don't like overindexing on eval behavior which I don't think is representative enough of real life behavior. To say nothing of goodhart problems if compliance incentives are highly focused on model cards.

Cite this project

@misc{lin2026automated,
  title = {{Automated Compliance Measurement for Frontier AI Models: Evidence-Based Scoring of Model Card Disclosures}},
  author = {Yulong Lin},
  year = {2026},
  month = feb,
  note = {Submitted to The Technical AI Governance Challenge, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/automated-compliance-measurement-for-frontier-ai-models-evidencebased-scoring-of-model-card-disclosures-4njc}},
  url = {https://apartresearch.com/sprints/projects/automated-compliance-measurement-for-frontier-ai-models-evidencebased-scoring-of-model-card-disclosures-4njc}
}

More from The Technical AI Governance Challenge

All 47 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026