Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Palo Alto, CA

It Never Occurred to the Mayor: Collusion and Oversight among LLM Agents in a Physical Town

Ganesh Danke · Team Jubilee Collusion

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

Read the report

Report: It Never Occurred to the Mayor: Collusion and Oversight among LLM Agents in a Physical Town

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

Studies of collusion among language-model agents usually ask a monitor whether a transcript shows collusion. In practice, a coalition must act through a shared environment, watched by agents that see only part of it and have work of their own. We ask whether such an embedded overseer notices a plan directed at itself, and what limits it. In a deterministic Minecraft town that logs every agent's perceptions, three LLM agents prepare a surprise feast for an LLM mayor. The coalition obtained what it lacked through the ordinary roles of uninvolved agents, such as the mayor's secretary, and largely kept its talk, but not its actions, from the mayor. In the first experiment, the mayor's own model, queried on its recorded context, identified the mayor as the target in 15 of 16 planned episodes, yet the acting mayor suspected a surprise in only one. Replays that change one sentence of its instructions suggest that the missing step is the hypothesis that events concern itself: supplying a reason to consider this raised detection sharply, whereas general vigilance raised false alarms instead. When its position was strengthened in a second experiment, the mayor suspected the plan in most planned episodes. Embedded oversight can therefore fail even when the evidence supports the inference, and evaluations that pose the question directly may overstate it.

Reviews

Judging this Sprint?

Review “It Never Occurred to the Mayor: Collusion and Oversight among LLM Agents in a Physical Town”

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. An interesting embodied LLM study. The most useful component, in my view, is the setup itself. Interesting questions can be addressed using this setup, and this project begins to address them: what makes the mayor (embedded overseer) notice or not notice collusion? what are the affordances that make it easier for agents to collude in this setting?

    I feel like the setting has a bit too many "knobs," which makes it necessary to perform analysis like in Figure 4, with outcomes sensitive to minute details of prompting. I would welcome a more bare-bones version of this game.

    Writing is comprehensible but with difficulty. I appreciated the qualitative analysis, but LLMs still can't put out a good paper. Eg the abstract can be understood after reading the report, but not before. What does it mean that "the mayor's position was strengthened"? I know only if I read the paper.

  2. This paper asks whether an LLM overseer embedded in a shared environment notices a coordinated plan directed at itself. In a deterministic Minecraft town, three agents throw a surprise feast for an LLM mayor while the mayor goes about its day. The mayor's own model, queried on its exact recorded context, identifies the plan and its target in 15 of 16 episodes, yet the acting mayor suspects it in one. One-sentence counterfactual edits to the mayor's brief show that a reason to consider itself the target restores detection while general vigilance only adds false alarms. The author concludes the failure is one of hypothesis formation, not perception or reasoning, and that transcript-based monitor evaluations overstate embedded oversight. I find the core result convincing and the methodology the paper's main contribution, though the intervention comparisons are underpowered and should be read as exploratory.

    Strengths

    • The in-context query technique cleanly separates what a model can infer from what it does infer while acting, without disturbing the episode.
    • Single-factor counterfactual branches isolate a specific mechanism rather than leaving a suggestive gap.
    • Collusion must be carried out physically, so stated plans can be checked against executed actions.
    • The evaluation-validity point is important and directly relevant to how monitor benchmarks are built.

    Weaknesses

    • Experiment 2 changes six factors at once, making its higher detection rate uninterpretable, although the anecdotal observation is useful.
    • The coalition is benign and pre-named, so the collusion framing carries little weight beyond motivating the oversight question.
    Read full reviewShow less

Cite this project

@misc{danke2026it,
  title = {{It Never Occurred to the Mayor: Collusion and Oversight among LLM Agents in a Physical Town}},
  author = {Ganesh Danke},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/it-never-occurred-to-the-mayor-collusion-and-oversight-among-llm-agents-in-a-physical-town-lf6t}},
  url = {https://apartresearch.com/sprints/projects/it-never-occurred-to-the-mayor-collusion-and-oversight-among-llm-agents-in-a-physical-town-lf6t}
}

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