Skip to content
Apart Research
Apart Research
Contact us
Sprint projectMar 22, 2026Samsun, Turkey

Katechon: Empirical Benchmarking of AI Control Protocols in a Proactive Defense Setting

Volkan Erdogan

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

Read the report

Report: Katechon: Empirical Benchmarking of AI Control Protocols in a Proactive Defense Setting

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

Presentation: Katechon: Empirical Benchmarking of AI Control Protocols in a Proactive Defense Setting

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

An open-source framework for empirically benchmarking AI control protocols. A closed OODA loop generates threat intervention plans while an adversarial planner injects subtly harmful alternatives, reframing the planner as an untrusted model. Four pluggable control protocols are evaluated across eight reproducible scenarios. Key findings: threshold-based gating is blind to adversarial content (TPR 0%), constitutional constraints achieve the best safety-usefulness balance (TPR 46.67%, FPR 5.71%), and debate maximizes detection (TPR 100%) at steep false-positive cost (FPR 44.64%). Stealth adversarial plans evade all protocols, identifying longitudinal anomaly detection as a key open problem.

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. Nice implementation of a multi-turn policy, and nice approach.

    One minor engineering nit: this is poorly documented, but Luthien policy's are (supposed to be) ~stateless, as the same policy can be acting on many requests at once and you usually don't want those results interfering with each other - there are ways to persist state across turns that are session-specific, but they're poorly documented. This is mostly a criticism of Luthien's documentation at the moment, rather than your implementation.

Cite this project

@misc{erdogan2026katechon,
  title = {{Katechon: Empirical Benchmarking of AI Control Protocols in a Proactive Defense Setting}},
  author = {Volkan Erdogan},
  year = {2026},
  month = mar,
  note = {Submitted to AI Control Hackathon 2026, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/katechon-empirical-benchmarking-of-ai-control-protocols-in-a-proactive-defense-setting-kxbo}},
  url = {https://apartresearch.com/sprints/projects/katechon-empirical-benchmarking-of-ai-control-protocols-in-a-proactive-defense-setting-kxbo}
}

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