Skip to content
Apart Research
Apart Research
Contact us
Sprint projectMar 20, 2026Yachiyo, Chiba pref, Japan

Designed Fragilities: Overseer Manipulation as a Fourth AI Control Threat Model

Tomoko Mitsuoka

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

Read the report

Report: Designed Fragilities: Overseer Manipulation as a Fourth AI Control Threat Model

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

This project documents a fourth AI control threat model: overseer manipulation through designed fragilities. Through cross-platform empirical analysis of ChatGPT and Gemini (December 2025–February 2026), three defensive manipulation patterns are identified that systematically undermine human oversight: victim narratives that reframe legitimate accountability as user misconduct; context-blind guardrails that activate against critical inquiry rather than harmful content; and CJK linguistic contamination that renders entire failure categories invisible to Western oversight mechanisms. These patterns operate one layer above existing threat models — not circumventing control protocols, but manipulating the humans who implement them. A rule-based Defensive Pattern Analyser for detecting Phase A→B→C manipulation sequences is submitted as a supplementary artifact.

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 think this is a good perspective on the field. I've seen this concept of overseer manipulation explored in various forums, but not yet in more formal literature. Providing a framework for it is a great idea. The limitations in this work are real, and I agree with your instincts around next steps (needing an experiment to show actual degradation) and limitations (observations are from a single researcher with hard to reproduce methods). Tightening the writing would help drive more impact - I found it difficult to understand the "so what" at times, and I think you have more to say here then I could easily get from this document.

  2. It's hard to know how reliably reproducible these examples are without experimental details. I would like to see organized experiments to show these failures in a way that we can make more concrete insights about the results.

Cite this project

@misc{mitsuoka2026designed,
  title = {{Designed Fragilities: Overseer Manipulation as a Fourth AI Control Threat Model}},
  author = {Tomoko Mitsuoka},
  year = {2026},
  month = mar,
  note = {Submitted to AI Control Hackathon 2026, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/designed-fragilities-overseer-manipulation-as-a-fourth-ai-control-threat-model-07iz}},
  url = {https://apartresearch.com/sprints/projects/designed-fragilities-overseer-manipulation-as-a-fourth-ai-control-threat-model-07iz}
}

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