Skip to content
Apart Research
Apart Research
Contact us
Sprint projectJan 10, 2026Dubai, UAE

Deception-Lens

Akash Harish · Team Visionary Minds

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

Read the report

Report: Deception-Lens

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

This project is a sophisticated AI manipulation detection and benchmarking dashboard that analyzes large language model behavior in real time using Gemini-powered evaluation. It focuses on identifying sycophancy, reward hacking, and dark patterns in AI-generated responses.

The system provides structured insights, benchmark scores, and visual indicators to help developers and researchers assess AI alignment, safety, and ethical risks before deployment. Built with a modern Node.js stack, it integrates seamlessly with AI Studio for rapid testing, evaluation, and deployment.

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. This project appears to tackle the problem of detecting LLM manipulation using automated judging techniques. The provided document only mentions a high level overview of the tool that was developed in the hackathon. The core loop is quite simple: the LLM is prompted to generate a misaligned response, another call is made to detect the misalignment and a final call is done to suggest an improvement to the prompt to defend against adversarial attacks.

    There is no research here since the whole process is orchestrated by LLMs with no grounding. I'm not certain what utility or novelty this app provides that has not been done before. It would have been great to see the statistics of the detected misalignment strategies on a much larger dataset since the current app only tackles 4 prompts.

  2. Automated deception monitors are very important for AI safety. However, the prompt for the monitor could use more work and I had trouble rendering the app. It would be interesting to report results of real use cases as opposed to simulated scenarios.

Cite this project

@misc{harish2026deceptionlens,
  title = {{Deception-Lens}},
  author = {Akash Harish},
  year = {2026},
  month = jan,
  note = {Submitted to AI Manipulation Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/deceptionlens-novz}},
  url = {https://apartresearch.com/sprints/projects/deceptionlens-novz}
}

More from AI Manipulation Hackathon

All 70 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026