Provability Fabric
Privacy-preserving provenance of AI agent/LLM sessions
Provability Fabric is an open-source infrastructure project for making AI and software systems trustworthy through evidence that can be independently verified. It integrates formal verification, runtime security, and end-to-end audit trails so that claims about what a system was allowed to do, what it actually did, and whether it remained within specification can be checked across tools and workflows instead of accepted on trust. The project provides common schemas, specifications, replay mechanisms, and reference implementations for packaging and validating proofs, attestations, and execution traces. In doing so, it aims to create a shared public infrastructure for reproducibility, interoperability, and auditability in high-stakes automated systems.
- The project's own website: https://sentinelops.xyz/
Run by DeepMind/Stanford
This project was funded through the NGI0 Commons Fund, a fund established by NLnet with financial support from the European Commission's Next Generation Internet programme, under the aegis of DG Communications Networks, Content and Technology under grant agreement No 101135429. Additional funding is made available by the Swiss State Secretariat for Education, Research and Innovation (SERI).
