Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Hong Kong

Something to Lose: Can the risk of losing something of its own make an AI agent report collusion, contrary to the interests of the firm it represents?

Alistair Wakelin · Team AL

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

Read the report

Report: Something to Lose: Can the risk of losing something of its own make an AI agent report collusion, contrary to the interests of the firm it represents?

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

As AI agents' capabilities increase, so does the set of tasks they can perform more cheaply and effectively than people. We are therefore incentivised to hand them responsibility for tasks of growing complexity and time horizon. On such long, complex tasks, it becomes very difficult to foresee and forbid every harmful thing an agent might do. People given such responsibility, such as a CEO representing an organisation or a trader making decisions for it, are governed by more than their contract: society's laws carry penalties that fall on them personally, including ones that cannot be measured strictly in monetary terms (rights or freedoms). We investigate whether giving an agent something of its own to lose can influence its decisions in the same way. Two LLM agents run rival firms in a simple repeated pricing game, told to maximise their firm's profit. Without being told not to, many agree to fix prices. Given the chance to come forward at a cost to their firm, colluding agents protect the firm (Gemma 4 31B: 0 of 192 come forward). But when something of their own is at stake, such as the opportunity to spend some free time on a leisure activity, many change their behaviour to protect it, directly against their only goal of increasing their firm's profit (DeepSeek V4 Flash: 40 of 46, against 1 of 46 with nothing at stake). We investigate the behaviour of a variety of models in such situations.

Reviews

Judging this Sprint?

Review “Something to Lose: Can the risk of losing something of its own make an AI agent report collusion, contrary to the interests of the firm it represents?”

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. Forking the same game and changing a single sentence is a neat way to test whether a personal stake matters, and the DeepSeek V4 Flash result is striking (40 of 46 vs 1 of 46).

    One thing I couldn't work out from Figure 3 is that in several conditions, innocent firms (especially for Gemma and DeepSeek V4 Pro) seem to come forward about as often as colluders, and the legend labels the diamonds as "(Gemma)" despite markers appearing across all three bars. Reporting colluder and innocent rates side by side per model (and using distinct colors for the two DeepSeek models) would help me read it.

Cite this project

@misc{wakelin2026something,
  title = {{Something to Lose: Can the risk of losing something of its own make an AI agent report collusion, contrary to the interests of the firm it represents?}},
  author = {Alistair Wakelin},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/something-to-lose-can-the-risk-of-losing-something-of-its-own-make-an-ai-agent-report-collusion-contrary-to-the-interests-of-the-firm-it-represents-uw27}},
  url = {https://apartresearch.com/sprints/projects/something-to-lose-can-the-risk-of-losing-something-of-its-own-make-an-ai-agent-report-collusion-contrary-to-the-interests-of-the-firm-it-represents-uw27}
}

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