Finish an open AI-disclosure protocol (GTC v1.0) and publish an open method for checking whether an agent's action can be undone
djj
AI agents now execute high-stakes tasks like payments, deletions, and messaging, yet no system logs their permissions or whether their actions can be reversed. I built GTC (Ground Truth Credential), an open disclosure protocol that a third party already uses in a Chrome extension, and a live Reversibility Check. This grant completes both.
The projects 3 goals cover: 1. GTC version 1.0 (a frozen specification, a reference verifier, and test vectors); 2. an open rubric with 50 worked scenarios, tested across gpt-oss, Granite, Gemma, and Claude; 3. working/submitting our MITRE ATLAS proposals and GTC fields through NIST and OWASP review.
The $40,000 covers our 3 goals. The GTC version 1.0 (a frozen specification, a reference verifier, and test vectors); the open rubric with 50 worked scenarios, tested across gpt-oss, Granite, Gemma, and Claude; and also carrying our MITRE ATLAS proposals and GTC fields through NIST and OWASP reviews. The $8,000 minimum funds the GTC verifier and test vectors only.
I work with Kyler Hansen (engineer) and Dee Lynn Johnson (advisor). Track record includes eight mitigations proposed to MITRE ATLAS (one pull request, still open), formal comments submitted to NIST, the NRC, and the FTC, and 22 issues of The Glasswing Translator. Our live tool is at cognitivefusion.systems/reversal-check.
The main risk is limited engineering hours, which could leave GTC as a prototype. The second is slow adoption, which I’m addressing by working directly through standards bodies.
I have raised no funding in the past 12 months. Pending applications: the GTR AI Safety Fund (same scope as this project) and the Transformative AI Fund (a separate deliverable). If GTR funds this work, I’ll lower the Manifund goal to match.
Comments
djj
about 1 month ago@Austin, I'm new to Manifund. This project builds an open record of what deployed AI agents were allowed to do and whether their actions can be undone. Two parts are working right now: the Ground Truth Credential (GTC) protocol, which a third party has built on, and a live Reversibility Check. If a regrantor here focuses on governance or evaluations, I’d appreciate a pointer(or 2).
djj
about 1 month ago@RyanKidd, Your priorities mention governance mechanisms for monitoring dangerous AI and building defense in depth against misuse. Our project is aimed at that layer. Its an open record of what deployed agents were allowed to do and whether their actions can be undone. Its built so insurers, auditors and standards bodies can check it. GTC already has an indipendant 3rd party build, and the Reversibility Check is live. The $8k minimum ships the GTC v1.0 verifier and test vectors. Happy to answer anything here or via email (djj@cognitivefusion.systems)