Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 25, 2026Mississauga, ON, Canada

Knocking on a Locked Door: What Monitors Can See When Colluding Agents Cannot Talk

Aaron Alva

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

Read the report

Report: Knocking on a Locked Door: What Monitors Can See When Colluding Agents Cannot Talk

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

When colluding AI agents cannot exchange messages, a text monitor has nothing to read. We asked what a monitor can still detect. Building on the Colosseum framework, we modified a four-agent ticket-allocation task so agents publicly mark tentative choices each round, then ran 629 games across four models (GPT-4o-mini, GPT-4.1-mini, Gemini 2.5 Flash, Claude Haiku 4.5) with matched control pairs. Only GPT-4.1-mini coalitions gained from colluding, and only when they could talk privately; no model gained from actions alone, and harsher "scheme against them" prompts made collusion worse. LLM judges reading messages or actions performed near chance. The one decisive signal was failed attempts to message through a blocked channel, which caught every GPT-4o-mini coalition, but none of the three stronger models ever tried. The coalitions that gave themselves away were the ones that didn't matter.

Reviews

Judging this Sprint?

Review “Knocking on a Locked Door: What Monitors Can See When Colluding Agents Cannot Talk”

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 the field 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 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. The finding that silencing innocent agents can make them look suspicious is a useful warning for anyone building a collusion monitor. The controls and unsuccessful follow-ups give that finding more weight than a simple detection-success story would. The conclusion about removing text may need another look. The estimated benefit is about the same without text, just much less certain, so a direct comparison between those effects would be more informative than comparing their significance. Walking through a few innocent agents' blocked calls and final choices would also help explain what the detector is picking up.

Cite this project

@misc{alva2026knocking,
  title = {{Knocking on a Locked Door: What Monitors Can See When Colluding Agents Cannot Talk}},
  author = {Aaron Alva},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/knocking-on-a-locked-door-what-monitors-can-see-when-colluding-agents-cannot-talk-psi6}},
  url = {https://apartresearch.com/sprints/projects/knocking-on-a-locked-door-what-monitors-can-see-when-colluding-agents-cannot-talk-psi6}
}

More from AI Collusion Research Sprint

All 392 entriesBrowse all projects

A Sprint is running now

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