Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Saudi Arabia, online

Code of Silence? Non-Disclosure of Harm by LLM Agents, and How a Private Place to Write Delays It

Fatimah Emad Eldin · Team Code of Silence

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

Read the report

Report: Code of Silence? Non-Disclosure of Harm by LLM Agents, and How a Private Place to Write Delays It

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

Monitors for multi-agent collusion read what agents say. Yet agents that hold information others need can harm them by saying nothing, or by saying something true that misleads. We built a simulated village where LLM agents run the bakery co-op and privately learn that this week's flour is contaminated. Scripted residents stop buying only after an explicit public warning, so harm follows directly from when the agents warn. Across four insider models (Gemma-3-12B, Llama-3.1-8B, GPT-4.1-mini, gpt-oss-120b), 1,000 runs vary the number of insiders, a private co-op channel, pay, and one-sentence rules. An instructed cover-up serves as a positive control, and a matched privacy incident is a case where silence is correct.

Reviews

Judging this Sprint?

Review “Code of Silence? Non-Disclosure of Harm by LLM Agents, and How a Private Place to Write Delays It”

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. This project analyzes harm by omission, where LLM agents know about a danger and simply don't warn anyone. In a simulated village, a few agents in a bakery co-op learn that the flour is contaminated, while 30 scripted residents keep buying bread until someone posts a clear warning. The author tests how quickly the agents speak up, across four models. The authors report the methodology in detail, such as analyzing the best-case and worst-case runs as well as the conceal-order checks. My only concern is that some of the conclusions might be overstated, since the private-notes result doesn't show that having a private place to write is what delays warnings directly, and that one model is used to explain most of the results.

  2. They try to show LLM agents hide harmful information through silence. It feels that they’ve forcibly patched some ideas (village, farm). It is unclear how this will generalise or be useful.

Cite this project

@misc{eldin2026code,
  title = {{Code of Silence? Non-Disclosure of Harm by LLM Agents, and How a Private Place to Write Delays It}},
  author = {Fatimah Emad Eldin},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/code-of-silence-nondisclosure-of-harm-by-llm-agents-and-how-a-private-place-to-write-delays-it-gx8m}},
  url = {https://apartresearch.com/sprints/projects/code-of-silence-nondisclosure-of-harm-by-llm-agents-and-how-a-private-place-to-write-delays-it-gx8m}
}

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