Skip to content
Apart Research
Apart Research
Contact us
Sprint projectMar 23, 2026New York City

Toolchain Adversaries: A Control Setting for Supply-Chain Attacks on AI Agent Pipelines

Alexander Liu

Submitted to AI Control Hackathon 2026. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Toolchain Adversaries: A Control Setting for Supply-Chain Attacks on AI Agent Pipelines

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

Supply-chain attacks on AI agent tool ecosystems have emerged as a practical threat, yet existing AI control evaluations assume the adversary is the model itself. We introduce a complementary control setting where the agent is always honest but a tool in its pipeline is compromised, enabling data exfiltration through the agent's faithful execution of tool outputs.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for AI safety 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 AI safety 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.

  1. Overall solidly motivated (this was written before LiteLLM supply chain attack). Supply chain attacks are a *known* issue, so I'm giving this a 4.5 instead of a 5. Execution and Presentation were about as good as I could have expected in a weekend. Overall this is about as high as I'm willing to give something that's essentially a solid proof of concept, and the author notes that this is different from a full-fledged experiment with more examples and ablations.

  2. The new control protocol and conceptual frame is interesting and worth developing. I'd like to see this extended.

Cite this project

@misc{liu2026toolchain,
  title = {{Toolchain Adversaries: A Control Setting for Supply-Chain Attacks on AI Agent Pipelines}},
  author = {Alexander Liu},
  year = {2026},
  month = mar,
  note = {Submitted to AI Control Hackathon 2026, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/toolchain-adversaries-a-control-setting-for-supplychain-attacks-on-ai-agent-pipelines-pb3g}},
  url = {https://apartresearch.com/sprints/projects/toolchain-adversaries-a-control-setting-for-supplychain-attacks-on-ai-agent-pipelines-pb3g}
}

More from AI Control Hackathon 2026

All 125 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026