Skip to content
Apart Research
Apart Research
Contact us
Sprint projectJul 27, 2026Sydney

Establishing a Framework for Analysing and Tracking Secret Loyalty Risk

Dhruv Hariharan

Submitted to Secret Loyalties Hackathon. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Establishing a Framework for Analysing and Tracking Secret Loyalty Risk

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

Track 5 This project takes the activation x action space, and: a) maps required capabilities to different regions b) identifies a series of risk models and maps them to the same space (with a slight treatment of AI coups) c) (briefly) quantifies the risk associated with each model (Table 1 summarises the previous points relatively briefly) d) makes some recommendations based on the results

(Apologies, I know the piece is very wordy - this is my first time writing a governance report, and I think I struggled with establishing a coherent argument, as opposed to just discussing some ideas. I think I went too overarching, and didn't do enough detailed work in analysing different risks. All feedback is greatly appreciated)

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. The scenario and the tables are good - it extends the framework and makes it more concrete in places that benefit from it. Taking this to a 4 or a 5 would require something in addition to this extension - e.g. a forecast or a fully developed worst case scenario for one of the rows.

    The table is excellent and is something the Kwon reference paper should have had in its original. It is solid analytical and thoughtful work, and I found it helpful. Some way to validate this would have gone a long way. E.g. fleshing out a full scenario for one of the rows or finding another way to validate these ideas.

    Writing is well organized and each scenario is well articulated and justified.

  2. The paper provides a competent high-level synthesis that maps capabilities and coup-adjacent scenarios onto Kwon et al.’s activation/action taxonomy and pairs it with a basic expected-damage decomposition, yet the exercise stays almost entirely qualitative and largely restates the existing research agenda without generating any testable prediction or empirical anchor.

    To raise its value, the author should replace the hand-wavy capability placements with concrete scores drawn from existing benchmarks (SHADE-Arena, SAD, CoT-red-handed) and run at least one minimal pilot that checks whether current models already exhibit the claimed “detection-risk estimation” or “concealment under monitoring” thresholds.

    As weekend threat modeling it is serviceable, but the absence of any validation or novel mechanism keeps it from being more than an incremental organizing note.

Cite this project

@misc{hariharan2026establishing,
  title = {{Establishing a Framework for Analysing and Tracking Secret Loyalty Risk}},
  author = {Dhruv Hariharan},
  year = {2026},
  month = jul,
  note = {Submitted to Secret Loyalties Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/establishing-a-framework-for-analysing-and-tracking-secret-loyalty-risk-dg45}},
  url = {https://apartresearch.com/sprints/projects/establishing-a-framework-for-analysing-and-tracking-secret-loyalty-risk-dg45}
}

More from Secret Loyalties Hackathon

All 177 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026