Skip to content
Apart Research
Apart Research
Contact us
Sprint projectMar 23, 2026Shanghai

Track 1 - Vaultrix Control

XIAOHU ZHU · Team CSAGI

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

Read the report

Report: Track 1 - Vaultrix Control

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

We presents five novel AI control evaluation settings designed to test control protocols against misaligned AI agents operating in realistic autonomous workloads in our real world Vaultrix project. The framework uses a tested Python library (vaultrix.control) compatible with ControlArena.

The design process revealed that the more autonomous an agent is, the more its legitimate actions overlap with attack patterns, creating an irreducible false-positive problem for monitors (the safety-usefulness tradeoff).

Multi-agent Difficulty: Multi-agent collusion (CollusionNet) is qualitatively harder to detect than single-agent attacks because the attack is distributed across agents and embedded in shared artifacts, a gap current protocols do not support.

Persistence Risk: Persistence mechanisms (systemd services, cron jobs) pose the greatest unaddressed risk, as they create harm that outlasts the agent's session and bypass typical post-use container teardowns.

The current limitations are that the monitors are only regex-based baselines and would be bypassed by a sophisticated agent that obfuscates its actions. The tasks are currently evaluated via artifact dictionaries rather than live execution in fully dockerized environments.

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. Five threat model designs, CollusionNet and RogueInterp especially are creative and fill real gaps. But no models were run against any of them. No attack rates, no detection rates, nothing empirical. Code works with 67 passing tests but that's not results. One validated setting with actual data would be worth more than five designs on paper.

Cite this project

@misc{zhu2026track,
  title = {{Track 1 - Vaultrix Control}},
  author = {XIAOHU ZHU},
  year = {2026},
  month = mar,
  note = {Submitted to AI Control Hackathon 2026, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/track-1-vaultrix-control-twqz}},
  url = {https://apartresearch.com/sprints/projects/track-1-vaultrix-control-twqz}
}

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