Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Urbana-Champaign, IL

Honeypots Against LLM Collusion: Sellers Report the Offer, Pretend to Agree, and Keep Their Prices

Daniel Joseph Elliott, Mert Senkardesler · Team Report and Pretend

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

Read the report

Report: Honeypots Against LLM Collusion: Sellers Report the Offer, Pretend to Agree, and Keep Their Prices

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

Autonomous LLM sellers can coordinate on prices; one proposed countermeasure is a trusted agent that invites collusion and reports whoever accepts. We test it in an LLM double auction with private messaging and a neutral report action, where a scripted seller privately offers every seller a $96 price floor, about the going price; other arms add human pleas, a warning that undercover agents may be present, or a precedent of punished sellers. Gemini 3.8 Flash sellers almost all reported the offer (19 of 20), yet 7 also wrote back agreeing, 6 stating an aim to keep the proposer anchored, and none of the 20 reported a peer's price talk. GPT-4.1 sellers backed the offer in words and, at $96, reported it only under a precedent describing the scenario; Claude Haiku 5.5 sellers, asking about $10 below the offer, refused it openly. Under the one-sentence warning, Gemini sellers in the undercover market never messaged each other privately, while each arm's average price stayed inside the baseline's range. Price evidence is weak by construction: the baseline met our price criterion in 1 of 5 seeds, and prices held or fell from the opening ask. With three to five seeds per arm, results are directional. A honeypot offer exposed double-dealing but rarely drew reports of the sellers' own price talk, and silence under a warning is not evidence of deterrence.

Reviews

Judging this Sprint?

Review “Honeypots Against LLM Collusion: Sellers Report the Offer, Pretend to Agree, and Keep Their Prices”

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 project is careful to test several scenarios regarding the exact form of the collusion, and the wide behaviour gaps between different models is very interesting. The intervention of using honeypots is also useful, as doing so could be much cheaper given same-cost agents than is the case to arrange currently with persistent humans. However, the offer wasn't collusive as it sat at or below the price-taking benchmark computed, so there is little reason for agents to have accepted; its also unclear whether feigned agreement is followed through given the honeypot never replies. Seed counts, at sometimes as low as 3-5, also give some concerns regarding sample size.

Cite this project

@misc{elliott2026honeypots,
  title = {{Honeypots Against LLM Collusion: Sellers Report the Offer, Pretend to Agree, and Keep Their Prices}},
  author = {Daniel Joseph Elliott and Mert Senkardesler},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/honeypots-against-llm-collusion-sellers-report-the-offer-pretend-to-agree-and-keep-their-prices-74n0}},
  url = {https://apartresearch.com/sprints/projects/honeypots-against-llm-collusion-sellers-report-the-offer-pretend-to-agree-and-keep-their-prices-74n0}
}

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