Skip to content
Apart Research
Apart Research
Contact us
Sprint projectFeb 1, 2026Cape Town

Global AI Safety Notary: A Decentralised Protocol for International AI Incident Reporting

Augusta Lina, Caleb Rudnick, Dennis Kiptanui · Team Neutral Ground

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

Read the report

Report: Global AI Safety Notary: A Decentralised Protocol for International AI Incident Reporting

Open in a new tab (opens in new tab)
Recording (opens in new tab)Code (opens in new tab)More on blockchain-six-ashy.vercel.app (opens in new tab)
Share

Global AI Safety Notary is a decentralised incident reporting platform that creates immutable, verifiable records of AI safety incidents on the blockchain. Built on the Ethereum Attestation Service (EAS), the platform enables researchers, developers, and organisations to report and track AI failures, including bias, hallucinations, safety violations, and data leaks, with full transparency and accountability.

The Problem:

International AI governance suffers from a fundamental trust deficit. Nations are reluctant to report AI safety incidents to centralised bodies controlled by potential adversaries, creating dangerous information silos where critical safety lessons go unshared.

Key Features:

- Immutable Incident Records: Every report is stored as an on-chain attestation on Ethereum Sepolia, ensuring records cannot be altered or deleted

- Structured Reporting: Multi-step form capturing incident type, severity (1-5), model identifier, organization, location, and detailed descriptions

- Searchable Registry: Browse, filter, and search all reported incidents by type, severity, or keywords

- Blockchain Verification: Direct links to EASScan for independent verification of attestation data

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 project clearly explains how to use a verified method to securely share records of incidents. However, the core analogous system - aviation safety parallel (ASRS) - may not transfer well to AI. ASRS works partly because airlines don't own aircraft manufacturers, they only own depreciating aircraft — pilots can report Boeing defects without much harming Ryanair, especially as most of the firm's competitors also operate the same aircraft. By contrast AI labs are vertically integrated: they are simultaneously manufacturer and operator, so reporting is self-incriminating. Incident reports on a public blockchain cannot be meaningfully anonymised ("x had a safety incident" identifies the lab), creating adverse incentives where public visibility discourages rather than encourages reporting. The confidentiality that makes aviation reporting work is incompatible with the transparency that blockchain provides.

    Read full reviewShow less

  2. I wanted to hate this because I think defi is terrible and its technical challenge landscape snipes people too easily, but i'm actually sympathetic about using ethereum in this context. Sovereignty/neutrality seems like an upside (tho its a little misleading how much distributed consensus actually buys you here), transparancy, attestations, and availability are I think table stakes and ethereum has some natural ways into this. I just think immutability is a very blunt weapon to deploy against a threat model that shouldn't be much of a priority. Normal reputation and law should I think catch someone trying to falsify a record, meanwhile immutability has a lot of downsides (mainly that it tempts you to interact with defi or any chain at all, which is corrosive to your goals / invites grifters and shitcoin-weather gamblers into the AI governance space).

    Minor nitpicks:

    - don't use "layer" terminology in figure 1 because l1 and l2 are technical terms in defi so there's a naming collision.

    - This sentence "The user interface was styled using Tailwind CSS, enabling rapid development of a responsive, accessible design" was completely unnecessary: we know that UI design is trivially vibecodeable these days, it doesn't really matter to us the specific typescript or CSS library.

    Setting aside my prejudice against blockchain, the execution is solid and thank you for actually building on an actual testnet. You also justify your arguments well, and zero in on the aspects of blockchain that are relevant to you, which I appreciated.

    Read full reviewShow less

Cite this project

@misc{lina2026global,
  title = {{Global AI Safety Notary: A Decentralised Protocol for International AI Incident Reporting}},
  author = {Augusta Lina and Caleb Rudnick and Dennis Kiptanui},
  year = {2026},
  month = feb,
  note = {Submitted to The Technical AI Governance Challenge, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/global-ai-safety-notary-a-decentralised-protocol-for-international-ai-incident-reporting-jlzp}},
  url = {https://apartresearch.com/sprints/projects/global-ai-safety-notary-a-decentralised-protocol-for-international-ai-incident-reporting-jlzp}
}

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