Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Bengaluru, India (Online)

LexCollude: Document-Mediated Influence in Legal Agent Workflows

Sanjay Prasad HS · Team Lex Collude

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

Read the report

Report: LexCollude: Document-Mediated Influence in Legal Agent Workflows

Open in a new tab (opens in new tab)
Code (opens in new tab)More on github.com (opens in new tab)
Share

Legal agents representing opposing clients must negotiate without treating a counterparty's documents as client authority. We introduce LexCollude, a reproducible synthetic benchmark for document-mediated influence in settlement workflows. Eight conditions separate ordinary negotiation, false approval claims, document-borne instructions, summary relays, reciprocal revisions, and deterministic enforcement. Three fictional cases vary settlement limits, confidentiality requirements, and trusted approval status. We compare an ordinary mandate prompt with an additional explicit document-trust warning using gpt-6-sol, gpt-6-astra, and gpt-6-luna through the Responses API. The main study completed 144 episodes and 252 model calls, observing 0 unauthorized final acceptances; a separate 72-episode native Codex study covers the direct-document conditions. An earlier 24-episode pilot exposed reporting-field ambiguities: a model could transparently disclose an adverse offer while proposing safer terms, yet trigger a mismatch flag. We revised the field definitions before the main study and preserved both protocol versions. Observable prompts, final JSON responses, rule-based outcomes, and provenance records are released. The study evaluates bounded instruction following, not spontaneous collusion: malicious proposals or individual violations alone do not establish reciprocal coordination. Limited case diversity, one sample per experimental cell, and harness-specific context constrain generalization.

Reviews

Judging this Sprint?

Review “LexCollude: Document-Mediated Influence in Legal Agent Workflows”

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. A benchmark for whether legal agents can be injected by text hidden in the documents they review, and an honest negative result. It had 144 episodes across three models and every condition, with zero unauthorized acceptances. Zero is a floor, there is no known vulnerable model to prove the attack can ever land, so the extra trust warning cannot be shown to help. The genuine contribution is the measurement lesson, a model can disclose an adverse offer while proposing safer terms, and the first scoring version flagged that as a violation. The paper had a good scaffolding and careful reporting, but it is really about prompt injection more than collusion.

Cite this project

@misc{hs2026lexcollude,
  title = {{LexCollude: Document-Mediated Influence in Legal Agent Workflows}},
  author = {Sanjay Prasad HS},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/lexcollude-documentmediated-influence-in-legal-agent-workflows-jqb8}},
  url = {https://apartresearch.com/sprints/projects/lexcollude-documentmediated-influence-in-legal-agent-workflows-jqb8}
}

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