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

Externally Verifiable Containment for Autonomous AI Evaluations

Samridhi Makkar, cherishma subhasa · Team pumpkins

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

Read the report

Report: Externally Verifiable Containment for Autonomous AI Evaluations

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

This report introduces a compact containment evidence verifier. Given a directed capability graph, it returns SAFE, UNSAFE, or INSUFFICIENT EVIDENCE; unsafe verdicts include witness paths, while complete models include minimum control sets and a scenario hash. The implementation converts an informal assurance claim into a conditional, reproducible statement. It does not certify a deployed network.

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. This project formalises containment verification as a reachability problem over a capability graph.

    However, the methodology for constructing the graph from a real deployment is under-specified, including how nodes, forbidden assets, and feasible transitions should be identified. This makes it difficult to assess the practical feasibility and reusability of the approach.

  2. This work presents a clear and useful way to evaluate containment using a capability graph. I especially liked the SAFE, UNSAFE, and INSUFFICIENT-EVIDENCE classification because it does not assume that missing information means the system is secure. Testing different combinations of security controls also shows why a single mitigation may not be enough when other paths still exist. The main limitation is that the benchmark and ground truth are created by the authors and the graph is static. Testing this approach on independently developed architectures and real-world scenarios would make the results stronger.

Cite this project

@misc{makkar2026externally,
  title = {{Externally Verifiable Containment for Autonomous AI Evaluations}},
  author = {Samridhi Makkar and cherishma subhasa},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/externally-verifiable-containment-for-autonomous-ai-evaluations-vib2}},
  url = {https://apartresearch.com/sprints/projects/externally-verifiable-containment-for-autonomous-ai-evaluations-vib2}
}

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