Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Darmstadt, Germany

A Game-Theoretic Bayesian Auditor for Detecting Collusion Between AI Agents

Kavita Kumari · Team Kavita

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

Read the report

Report: A Game-Theoretic Bayesian Auditor for Detecting Collusion Between AI Agents

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

AI agents now set prices for real businesses, and two agents in the same market can work together to keep prices high. An auditor usually sees only the prices, not what the agents say to each other. One of the existing tools, pairwise net influence, looks for agents whose prices are linked. I show that this misses a whole kind of collusion: two agents who simply agree never to charge a low price, which leaves no link between their actions, only a shift in each one's behavior. I model detection as a game between a hidden colluding pair and an auditor, and I also build a Bayesian auditor that checks both traces and accuses a pair only when it is at least 95% sure. In a four-firm pricing game with Llama-3.3-70B agents, my auditor identified the correct pair in all 12 colluding markets, with no false alarms. Detectors (net influence, MI test, price correlation, and an LLM judge ) that look only at information between agents missed every market where the colluders simply avoided low prices.

Reviews

Judging this Sprint?

Review “A Game-Theoretic Bayesian Auditor for Detecting Collusion Between AI 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. This is a strong paper with a clear approach to detecting collusion between AI agents. The key insight, that collusion leaves two kinds of clues and older detectors only look for one, is simple and convincing. The experiments are thorough, and it is impressive that the results match the predictions so closely. One obvious next step is to add more test runs and more LLM models. Another important area is false positives, i.e., how to tell an honest firm that simply behaves differently apart from a real colluder. Also, the auditor knows the true honest behaviour, but the other detectors do not have that information, so giving them the same information would make the comparison fairer.

Cite this project

@misc{kumari2026gametheoretic,
  title = {{A Game-Theoretic Bayesian Auditor for Detecting Collusion Between AI Agents}},
  author = {Kavita Kumari},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/a-gametheoretic-bayesian-auditor-for-detecting-collusion-between-ai-agents-5ou9}},
  url = {https://apartresearch.com/sprints/projects/a-gametheoretic-bayesian-auditor-for-detecting-collusion-between-ai-agents-5ou9}
}

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