Skip to content
Apart Research
Apart Research
Contact us
Sprint projectSep 14, 2026Shenzhen

From Loss of Containment to Regulatory Inquiry-An Article 91 Information-Request Framework for Frontier AI Incidents

Rosie · Team Tracebound

Submitted to AI Incident Response Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: From Loss of Containment to Regulatory Inquiry-An Article 91 Information-Request Framework for Frontier AI Incidents

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

This project develops an **Incident-to-Inquiry Mapping** framework for regulatory response to frontier AI incidents. Using the 2026 OpenAI–Hugging Face incident as a case study, it maps public evidence to legal uncertainties, relevant EU AI Act provisions, and targeted information requests under Article 91. The project also proposes a stepwise escalation pathway from fact-finding to Article 92 evaluation and, where justified, Article 93 mitigation measures.

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 appreciate the efforts to systematically break an incident down into actionable questions that a regulator could ask. I would like to see more effort put into 'related work,' as I believe there are numerous frameworks for dissecting incidents.

  2. The paper states that it maps each question to a specific provision, and it would benefit from showing more clearly that it does so. Five of the six rows in the table in 4.1 hook onto Article 55(1)(a)–(d), while Article 53 and Annex XI do not appear in the paper at all, so the question of how much of the requested evidence is already required as part of the technical documentation is not asked.

    The Code of Practice is listed as a reference but is not really used in the analysis: there is no discussion of the Safety and Security Framework or of the model reports that OpenAI, as a signatory, submits to the AI Office, and Parts C and D of the appendix leave the Commission's reporting template for serious incidents unaddressed. Overall the paper would benefit from a much deeper legal analysis of what is already covered by the Code or by Article 53 anyway.

    Some of the controls also seem to restate existing law. C4, the minimisation control, is arguably what the EU principle of proportionality already requires, and C5, together with the claim in 5.3 that the framework "disciplines escalation", describes the sequence the AI Act itself sets out. Article 91 is likewise treated as the default information-gathering step without any discussion of informal requests.

    Finally, the method rests on public incident evidence - the stated aim is to show "how the EU AI Office could convert public incident evidence into a proportionate, legally grounded request for information" - but what if there is none? Relying on what providers have chosen to disclose raises the question of why they would, and what the incentives are, which is precisely the gap a regulatory instrument should close. Focusing on that question could make the paper considerably stronger.

    Read full reviewShow less

Cite this project

@misc{rosie2026from,
  title = {{From Loss of Containment to Regulatory Inquiry-An Article 91 Information-Request Framework for Frontier AI Incidents}},
  author = {Rosie},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/from-loss-of-containment-to-regulatory-inquiryan-article-91-informationrequest-framework-for-frontier-ai-incidents-c40t}},
  url = {https://apartresearch.com/sprints/projects/from-loss-of-containment-to-regulatory-inquiryan-article-91-informationrequest-framework-for-frontier-ai-incidents-c40t}
}

More from AI Incident Response Sprint

All 233 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026