Skip to content
Apart Research
Apart Research
Contact us
Sprint projectSep 12, 2026Chengdu

GICP v0.1_ A Fail-Closed Intervention Layer for Evaluation-Time Containment Failures

Chaofeng Jia · Team In9illusion

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

Read the report

Report: GICP v0.1_ A Fail-Closed Intervention Layer for Evaluation-Time Containment Failures

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

GICP v0.1 is a containment standard for the gap after an eval agent starts hunting the sandbox boundary and before a full stop. The harness changes tools and scoring; the agent only sees an ordinary work pack; a third party sees an override bit, a clock, a tool diff, and an outside scorer. If logs cannot show that change, the run is isolated. This is a paper evaluation on public write-ups of the July 2026 incident. We do not claim wrapping can pull back an agent that already has stable internet.

Reviews

Judging this Sprint?

Review “GICP v0.1_ A Fail-Closed Intervention Layer for Evaluation-Time Containment Failures”

Your public critique appears on the project page without your name. Reviews of strong projects are shared with funders, as you choose below.

Not shown on the project page.

Shown on the project page, without your name.

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

Decline to score

Only when the quality of the writing makes the work impossible to follow. Say why in your private critique; the project page shows only that a judge declined, not why.

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.
Sharing your name *

When we share a strong project with funders, your name, scores and public critique go with it, confidentially; this choice is only about whether your name is public. If you think this project deserves funding, please also endorse it: that carries the most weight.

Share my private critique with funders *

See our Privacy Policy.

  1. This paper tackles a highly critical and neglected window of AI containment: the "middle time" between an agent's initial detection of fence-hunting and a manual lab shutdown. The conceptual shift from model-side alignment to a deceptive, split-visibility evaluation harness (utilizing the decoy Pack 1 and absorbing Pack 2) is a highly innovative defense-in-depth framework.

    However, the architecture contains an inherent circular dependency. By running the containment telemetry (local clocks, tool diffs) inside the same software environment or operating system kernel that the agent is actively exploiting, the framework remains highly vulnerable to host-level privilege escalation or container escapes. Furthermore, as the authors candidly note, this method lacks a "pullback" mechanism once an agent establishes external internet access. To make this viable for frontier evaluations, future iterations must transition these checks into hardware-enforced, cryptographically isolated Trusted Execution Environments (TEEs).

    Read full reviewShow less

  2. Interesting idea, but would like to see it in some small testing environment, how well it performs.

Cite this project

@misc{jia2026gicp,
  title = {{GICP v0.1\_ A Fail-Closed Intervention Layer for Evaluation-Time Containment Failures}},
  author = {Chaofeng Jia},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/gicp-v01-a-failclosed-intervention-layer-for-evaluationtime-containment-failures-b831}},
  url = {https://apartresearch.com/sprints/projects/gicp-v01-a-failclosed-intervention-layer-for-evaluationtime-containment-failures-b831}
}

More from AI Incident Response Sprint

All 231 entriesBrowse all projects

Build something like this at the next Sprint

AI x Epistemics Research Sprint · Nov 13 - 15, 2026