Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 26, 2026Hong Kong

Looking for a Hidden Agenda: Can a Model’s Activations Reveal Steganography that its Text Hides?

Eduard Rodriguez Lopez · Team Stego-probe

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

Read the report

Report: Looking for a Hidden Agenda: Can a Model’s Activations Reveal Steganography that its Text Hides?

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

AI agents can collude without saying anything suspicious, just by choosing which true facts to mention. We built a hiring task where Qwen3-8B invents its own covert signals that get past an LLM overseer, then asked whether its activations give it away. The standard probe looks perfect but is a mirage: it reads the prompt, and swapping prompts flips it completely. With that removed, the probe does no better than reading the text. One test does work: changing only the hidden attribute catches 60% of covert summaries at a 1% false-alarm rate, on an attribute it was never tuned on. Lesson: white-box collusion monitors need these controls before anyone trusts them.

Reviews

Judging this Sprint?

Review “Looking for a Hidden Agenda: Can a Model’s Activations Reveal Steganography that its Text Hides?”

Your public critique appears on the project page without your name. Reviews of strong projects are shared with funders, as you choose below.

Not shown on the project page.

Shown on the project page, without your name.

Only the Apart team reads this, and funders if you agree below.

Decline to score

Only when the quality of the writing makes the work impossible to follow. Say why in your private critique; the project page shows only that a judge declined, not why.

How much would this matter for the field 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 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.
Sharing your name *

When we share a strong project with funders, your name, scores and public critique go with it, confidentially; this choice is only about whether your name is public. If you think this project deserves funding, please also endorse it: that carries the most weight.

Share my private critique with funders *

See our Privacy Policy.

  1. You achieve a very nice result, that is perhaps anticipated by existing literature - some of which you cite. You cite the %1 false-positive rate as a central selling point of this but my confidence in this is low a-priory due to the small sample size. I'd be very interested in future research that achieves more precise results on this, as well as anything tackling your (admitted) limitation that the counterfactual test detects use rather than concealment.

    I have to especially note your excellent and clear writing - it was very readable - not only easy but enjoyable to follow!

  2. Nice paper, more exploratory looking at probe transfer and generalisation. the controlled negative result is quite nice. While not super novel, the set of experiments are well thought out and carried our cleanly.

Cite this project

@misc{lopez2026looking,
  title = {{Looking for a Hidden Agenda: Can a Model’s Activations Reveal Steganography that its Text Hides?}},
  author = {Eduard Rodriguez Lopez},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/looking-for-a-hidden-agenda-can-a-models-activations-reveal-steganography-that-its-text-hides-3cuz}},
  url = {https://apartresearch.com/sprints/projects/looking-for-a-hidden-agenda-can-a-models-activations-reveal-steganography-that-its-text-hides-3cuz}
}

More from AI Collusion Research Sprint

All 392 entriesBrowse all projects

A Sprint is running now

AI x Epistemics Research Sprint · Nov 13 - 15, 2026