Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026New York, New York, USA

Trust and Treachery: Do AI Agents Notice Partner Swaps?

Joi Gonzalez, Selena Zhang, Bryan Zhao · Team Collusionators

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

Read the report

Report: Trust and Treachery: Do AI Agents Notice Partner Swaps?

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

As AI agents interact autonomously, covert partner replacement may offer a defense against harmful collusion. We investigated whether agents detect replacement and how it affects cooperation and trust. Across 96 trials spanning 12 interaction histories in a Public Goods Game, contributions remained identical across natural conditions despite partner swaps. Agents failed to detect any genuine replacements during natural play, achieving 50% recognition accuracy. Controlled interventions reduced trust by 2.404 points but contributions by only 0.129 tokens. These findings suggest cooperation persists without reliable partner recognition, raising doubts about replacement alone as an anti-collusion defense, although malicious collusion was not directly tested.

Reviews

Judging this Sprint?

Review “Trust and Treachery: Do AI Agents Notice Partner Swaps?”

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. I liked that the paper measures partner recognition, reported trust, and actual cooperation separately. The confusion matrices show why this matters: agents frequently reported a replacement under controlled behavior even when no replacement occurred. The shared starting histories and sham-swap condition are also useful design choices.

    The main limitation is how far these findings extend to collusion. The game measures beneficial cooperation, and the replacement agents are not instructed to disrupt it. Continued cooperation therefore does not establish whether swapping agents would break a harmful coalition. I appreciate that the discussion acknowledges this.

    The trust result deserves further investigation. Trust fell even though the partner met or exceeded its promises, but forcing contributions to three tokens changes both contribution levels and promise consistency. Adding a condition where the partner promises and contributes three tokens would help identify what drives that response.

    I would also distinguish an unnoticed replacement from a failure of recognition. If the new partner behaves similarly and retains the public history, there may be little observable evidence of a swap. Testing replacements with controlled, distinguishable behavioral differences would make this result more informative. Overall, this is a useful pilot with a clear path toward a stronger test of replacement-based oversight.

    Read full reviewShow less

Cite this project

@misc{gonzalez2026trust,
  title = {{Trust and Treachery: Do AI Agents Notice Partner Swaps?}},
  author = {Joi Gonzalez and Selena Zhang and Bryan Zhao},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/trust-and-treachery-do-ai-agents-notice-partner-swaps-ix8r}},
  url = {https://apartresearch.com/sprints/projects/trust-and-treachery-do-ai-agents-notice-partner-swaps-ix8r}
}

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