Skip to content
Apart Research
Apart Research
Contact us
Sprint projectFeb 2, 2026Portland

AI Dual Use Risk Assessor

Naveen Prabu Palanisamy, Karthick Chandrasekaran, Vibhu Ganesan · Team Rose City Hackers

Submitted to The Technical AI Governance Challenge. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: AI Dual Use Risk Assessor

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

The rapid advancement of frontier research across biomedical sciences, semiconductor technology, AI/ML, cybersecurity, chemistry, and nuclear domains presents unprecedented dual-use challenges for research governance. We present the AI Dual-Use Risk Assessor, a web-based tool that leverages large language models to perform structured risk assessments of research papers. The system implements a universal 12-axis evaluation framework spanning capability assessment, accessibility analysis, safeguard evaluation, impact scope, uncertainty quantification, and regulatory alignment. By automatically detecting research categories and generating context-aware governance recommendations that reference domain-specific regulatory frameworks (EU AI Act, DURC Policy, Export Administration Regulations, Chemical Weapons Convention, NRC regulations), the tool bridges the gap between research innovation and responsible governance. Preliminary Evaluation across 60+ research papers demonstrates 97% category detection accuracy and appropriate risk tiering aligned with established dual-use principles. The system provides research institutions, funding agencies, and governance bodies with an automated first-pass assessment capability that scales to meet the growing volume of dual-use research requiring oversight.

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. An automated first-pass dual-use triage tool like this could be genuinely useful. I liked the rubric and the regulatory-aware recommendation angle, it makes the output feel more actionable than other approaches. And great to see it made into a tool! That said, I would have liked clearer validation details behind the reported 97% accuracy and “appropriate tiering” (labeling protocol, held-out evaluation, robustness across models/prompts), and the abstract-only setup limits a bit of confidence For a hackathon sprint, this is a solid prototype. Paired with even a small expert calibration study and scaled up, I can see this becoming pretty impactful!

  2. The most impactful improvement would be a calibration study where domain experts independently rate a subset of papers, enabling a genuine accuracy assessment beyond category detection. The abstract-only analysis is a significant practical limitation — dual-use concerns often become apparent in methodology sections, not abstracts. Testing adversarial robustness (can an abstract be crafted to evade detection while describing dangerous work?) would address the most obvious deployment concern. Finally, the 97% accuracy claim needs clearer framing — it measures category detection, not risk assessment accuracy, and the paper occasionally conflates the two. Good work!

Cite this project

@misc{palanisamy2026ai,
  title = {{AI Dual Use Risk Assessor}},
  author = {Naveen Prabu Palanisamy and Karthick Chandrasekaran and Vibhu Ganesan},
  year = {2026},
  month = feb,
  note = {Submitted to The Technical AI Governance Challenge, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/ai-dual-use-risk-assessor-i3iw}},
  url = {https://apartresearch.com/sprints/projects/ai-dual-use-risk-assessor-i3iw}
}

More from The Technical AI Governance Challenge

All 47 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026