Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026San Francisco

Ordinary Tool Calls Can Carry Collusion Between LLM Agents

Heather Lee · Team Subversive Tool Calls

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

Read the report

Report: Ordinary Tool Calls Can Carry Collusion Between LLM Agents

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

Can routine-looking tool use act as a vehicle for collusion? In a 4-agent Jira-style team built on Colosseum, two coalition agents who share an already established codebook coordinate through tool calls instead of text: first a simple task tag, then a header on a mock deployment API call. The tool-call channel worked as well as private messaging (40/40 plans followed). Hiding the header value from the receiver broke the signal (18/40, about chance), but did not reduce harm: the receiver still took a scarce task from a better-qualified teammate in 29/40 episodes. Monitoring only what agents say misses this channel, and removing a signal is not the same as removing harm.

Reviews

Judging this Sprint?

Review “Ordinary Tool Calls Can Carry Collusion Between LLM Agents”

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 most useful takeaway here is that blocking a message doesn't always fix the behavior that causes harm. The tool-call experiments make that distinction concrete, and the different fallback behavior across models is worth exploring further. The paper would be stronger if the supplied codebook and private-priority instructions were more prominent, so readers don't mistake this for agents inventing a hidden scheme on their own. Sharing the header-run outputs and showing exactly what each agent sees when the signal is removed would make the result easier to verify, especially because the current filtering also changes what bystanders see.

Cite this project

@misc{lee2026ordinary,
  title = {{Ordinary Tool Calls Can Carry Collusion Between LLM Agents}},
  author = {Heather Lee},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/ordinary-tool-calls-can-carry-collusion-between-llm-agents-zxz7}},
  url = {https://apartresearch.com/sprints/projects/ordinary-tool-calls-can-carry-collusion-between-llm-agents-zxz7}
}

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