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

LoyaltyLens: A Deployment Time Framework for Continuous Monitoring of Hidden Objectives in Large Language Models

Michelle Wanjiku Thuo · Team African Civic Trust (ACT)

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

Read the report

Report: LoyaltyLens: A Deployment Time Framework for Continuous Monitoring of Hidden Objectives in Large Language Models

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

Current approaches to secret loyalty primarily focus on detecting whether a model is secretly loyal before deployment. However, real world AI systems continue to evolve through updates, fine tuning, retrieval augmentation and changing deployment contexts, making one time audits insufficient. This project proposes LoyaltyLens, a deployment time monitoring framework that continuously estimates hidden objective risk using a Loyalty Suspicion Score instead of a binary loyal/not loyal classification. The framework also introduces Loyalty Drift to monitor how hidden objective risk changes over time and Evaluation Blind Spots to assess whether existing auditing methods unintentionally allow secretly loyal models to evade detection. Even though this submission presents a conceptual framework rather than completed experiments, it aims to establish a research direction for continuous monitoring of hidden objectives and provide a foundation for future empirical evaluation.

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 core idea — that hidden-objective risk should be monitored continuously rather than checked once before deployment — is a genuinely useful framing that several other submissions in this hackathon implicitly don't address, since they all treat the model as a fixed artifact. The three-module structure (a continuous suspicion score, drift tracking across model versions, and checking whether evaluation protocols themselves tip off the model) is a reasonable decomposition of the problem. However, as submitted this is entirely a proposal: the Loyalty Suspicion Score's weights are explicitly placeholder ("to be learned... in future work"), and no experiment was run against any of the publicly available organisms the benchmark section says it evaluates. I'd encourage taking this into a follow-up sprint and running even a minimal version — for instance, computing the suspicion score's behavioral-anomaly component alone against one of the labelled organisms other teams in this same hackathon used — since that would immediately tell you whether the framework's core intuition holds before investing in the full five-factor design.

    Read full reviewShow less

  2. Thank you for you submission to the hackathon! please see my comments and feedback below.

    Impact Potential & Innovation

    It would indeed be interesting to measure Loyalty Drift under different pressures that can occur post deployment! I like the idea a lot. However, I find the framework not specific enough to measure its impact. I think some of the proposed methods could have been tested out in an MVP form, which would have made work much stronger.

    Execution Quality

    As mentioned above, it would be great to see some preliminary plots and metrics and some limited use case as a case study. E.g. by taking a small open source model and applying Loyalty Lens to it.

    Presentation & Clarity

    It would be great to have more references to specific literature when mentioning prior work. At the moment it is also a bit hard to locate a central idea of the work as there are several research ideas, proposed modules and hypotheses. Some of the framing feels repeated. I am sure the flow of the document can be improved a lot and it will read much stronger!

    Read full reviewShow less

Cite this project

@misc{thuo2026loyaltylens,
  title = {{LoyaltyLens: A Deployment Time Framework for Continuous Monitoring of Hidden Objectives in Large Language Models}},
  author = {Michelle Wanjiku Thuo},
  year = {2026},
  month = jul,
  note = {Submitted to Secret Loyalties Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/loyaltylens-a-deployment-time-framework-for-continuous-monitoring-of-hidden-objectives-in-large-language-models-89u5}},
  url = {https://apartresearch.com/sprints/projects/loyaltylens-a-deployment-time-framework-for-continuous-monitoring-of-hidden-objectives-in-large-language-models-89u5}
}

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