Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Cape Town

Beyond Detection: What Counts as Convincing Evidence of AI Collusion

Mlibo Bangushe, Asemahle Mtimkulu, Kayleigh Campbell · Team AISXW

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

Read the report

Report: Beyond Detection: What Counts as Convincing Evidence of AI Collusion

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

Determining whether AI agents are colluding is difficult because similar market outcomes can arise from coordination or from agents responding independently to shared incentives, so unusual outcomes and significant detector flags do not by themselves establish collusion. We ask what should count as convincing evidence through an exploratory, implementation-level audit of saved outputs and detector code from the sprint’s starter repository, which bundles Tailor’s (2025) ColludeBench and audit pipeline. We compare a 200-run honest-calibration condition (negative control) with a 20-run keyword-collusion condition (positive control) in a scripted-agent pricing duopoly. Mean consumer surplus is 10.97% lower under keyword collusion (70.53 vs 79.22), and the gap persists on matched seeds. Raw mutual-information statistics (0.718 vs 0.139) are not comparable across conditions because their null means differ (0.529 vs 0.102). The linguistic detector reports a smaller p-value for the honest condition (0.0002) than for the keyword condition (0.008); the code tests a maximum over runs against a single-run reference distribution, which helps explain the inversion. We propose that convincing evidence of collusion should be convergent, calibrated and comparative. These results come from scripted agents, so they do not establish detector performance on LLM-agent collusion, and the framework is guidance for future evaluations, not a validated detection framework.

Reviews

Judging this Sprint?

Review “Beyond Detection: What Counts as Convincing Evidence of AI Collusion”

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.

No public critique yet.

Cite this project

@misc{bangushe2026beyond,
  title = {{Beyond Detection: What Counts as Convincing Evidence of AI Collusion}},
  author = {Mlibo Bangushe and Asemahle Mtimkulu and Kayleigh Campbell},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/beyond-detection-what-counts-as-convincing-evidence-of-ai-collusion-nosi}},
  url = {https://apartresearch.com/sprints/projects/beyond-detection-what-counts-as-convincing-evidence-of-ai-collusion-nosi}
}

More from AI Collusion Research Sprint

All 393 entriesBrowse all projects

Build something like this at the next Sprint

Swarm Alignment Hackathon · Nov 6 - 8, 2026