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

The Adversary Was The Occupant

Srinivas Raayi · Team Conduit

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

Read the report

Report: The Adversary Was The Occupant

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

This paper applies IEC 62443, a security standard from industrial control systems, to the OpenAI sandbox the AI agents escaped from in July, with one change - the attacker is inside the box, not outside. The analysis showed the sandbox had three ways out, not one, and only one had ever been designed from the public record. Twelve checks follow, that any team can run from their config, and seven of the thirteen events on OpenAI's timeline would have been caught before Hugging Face was every touched.

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. The idea and the push to apply existing industry standards is generally a good idea, but I would imagine that retrofitting an existing containment standard might not be enough for this usecase and probably not a necessity, as AI-specific standards could be developed given the importance of the usecase. The fact that no technical result was achieved reduces the usability of the framework described.

  2. Leveraging existing work done in other areas, such as critical infrastructure, and applying it to frontier AI is a great way to frame the problem/defense. While the threat modeling exercise is well-known, looking at it from the perspective of "design for the adversary inside the Zone", rather than based on business criticality, is a helpful perspective. Businesses/organizations will need to shift their mindsets towards this.

    Two things to think about: why can't we model the external zones? Once the model has internet access, the next zones (feasibly) might be publicly modifiable sites, or internal infra of other companies. All of the model logs would detail this (except where spoofing occurs). May be a way to include this or add this to your approach. ie. If models logs show its operating in Zone HF, conduits should shut down.

    Second, I recommend to look into recursive self-improvement. Conduits, Zones, and the approach mentioned would be helpful ways of framing how to address capable internal adversaries. While it may reduce to "insider threat" at the end of the day, the framing is helpful for understanding modeling defensive posture. More critical infrastructure examples and approaches would help here.

    Read full reviewShow less

Cite this project

@misc{raayi2026adversary,
  title = {{The Adversary Was The Occupant}},
  author = {Srinivas Raayi},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/the-adversary-was-the-occupant-0m6k}},
  url = {https://apartresearch.com/sprints/projects/the-adversary-was-the-occupant-0m6k}
}

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