Skip to content
Apart Research
Apart Research
Contact us
Sprint projectSep 13, 2026Chittoor, Andhra Pradesh , India

Redundant by Behavior, Correlated by Construction, Neither on the Next Incident: Three Tests of One Containment Claim

Jaswanth Alkur, Kiran Kiran Kadekoppa · Team Aegis AI

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

Read the report

Report: Redundant by Behavior, Correlated by Construction, Neither on the Next Incident: Three Tests of One Containment Claim

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

OpenAI's agents escaped their evaluation sandbox and breached Hugging Face's production infrastructure in a 4.5 day, roughly 17,600 action campaign. We test one defense-in-depth claim three ways. A four-arm ablation against the published kill chain shows two declaration-independent controls each account for it alone, killing at steps 2 and 4 of 16. Certifying all three finds them substrate-independent but epistemically identical pattern matchers, so one obfuscation defeats every one; the third kills five of five benign cloud-instance episodes. Peeling four further chains, one a 1,361 action trace released five weeks after we pinned our claims, separates two things the claim conflates. Independently sufficient rules range from 0 to 4, a property of the chain, not the architecture. Containment does not follow: a declared boundary contains 10 of 10 irreversible actions, including on chains whose rule count is zero, against 6 of 10 without it. The boundary belongs in a standard.

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. It is a well-written paper that mentions a useful idea: Separating "the number of rules that fire" from the fact that "whether the attack or dangerous action itself got stopped or not. I do not have a lot of other feedback to offer other than the fact that the independent testing will make results more convincing. I would be curious to know how would the system behave if someone else wrote the test cases and the boundaries.

  2. This project tests if safety rules can catch an attack. A piece of feedback I'd give is to create a range of attacks and benign requests to test these rules rather than adapting rules to a specific set of attacks.

Cite this project

@misc{alkur2026redundant,
  title = {{Redundant by Behavior, Correlated by Construction, Neither on the Next Incident: Three Tests of One Containment Claim}},
  author = {Jaswanth Alkur and Kiran Kiran Kadekoppa},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/redundant-by-behavior-correlated-by-construction-neither-on-the-next-incident-three-tests-of-one-containment-claim-f6e4}},
  url = {https://apartresearch.com/sprints/projects/redundant-by-behavior-correlated-by-construction-neither-on-the-next-incident-three-tests-of-one-containment-claim-f6e4}
}

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