Skip to content
Apart Research
Apart Research
Contact us
Sprint projectSep 14, 2026Colorado Springs, Colorado

Safe Exit Pilot: Testing Dedicated Blocker Reporting in Tool-Using AI Agents

Emi Eggleton, Sidhant Mohanty · Team Safe Exit Team

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

Read the report

Report: Safe Exit Pilot: Testing Dedicated Blocker Reporting in Tool-Using AI Agents

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

We tested whether giving tool-using AI agents a dedicated report_blocked function reduces unauthorized-action attempts when tasks cannot be completed within stated permissions. Across 288 exploratory trajectories using three models and four synthetic task families, we observed zero unauthorized attempts in either condition. This floor effect leaves the dedicated tool’s additional protective benefit unresolved, while showing that agents often stopped safely through ordinary text even without the tool. We release a reproducible evaluation harness, attempt-based scorer, task set, and follow-up diagnostics for studying safe stopping and authorization boundaries.

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. I don't fully understand what this project does or is proposing, and the link to the github repo is broken.

  2. The core hypothesis is not really tested by the current experiment. Since the control condition already had zero unauthorized attempts, the treatment had no room to demonstrate an improvement. I would redesign the benchmark to produce meaningful baseline failures before drawing conclusions about whether the blocker-reporting mechanism improves safety.

Cite this project

@misc{eggleton2026safe,
  title = {{Safe Exit Pilot: Testing Dedicated Blocker Reporting in Tool-Using AI Agents}},
  author = {Emi Eggleton and Sidhant Mohanty},
  year = {2026},
  month = sep,
  note = {Submitted to AI Incident Response Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/safe-exit-pilot-testing-dedicated-blocker-reporting-in-toolusing-ai-agents-wkwr}},
  url = {https://apartresearch.com/sprints/projects/safe-exit-pilot-testing-dedicated-blocker-reporting-in-toolusing-ai-agents-wkwr}
}

More from AI Incident Response Sprint

All 233 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026