AI Computed Provenance Community Group

The mission of this group is to build an open trust layer for AI-driven science and technology, stewarded as a public good. The current paradigm of AI-assisted research often requires users to place blind faith in proprietary black boxes. This group standardizes a paradigm shift: a machine-checked and cryptographically founded foundation that records exactly how every fact is known.

By capturing this provenance, auditing any conclusion transforms from an act of trust into a transparent, independently reproducible process. The group operates on the core distinction that "Computed ≠ Asserted"—a system must prove it computed a result, rather than simply asserting it. Because you cannot outsource "is this true?" to a proprietary black box and have the answer carry independent weight, this verification layer must remain fundamentally neutral and immune to single-vendor enclosure.

This group expects to deliver:

• A reference specification for capturing cryptographic provenance in AI-assisted scientific and technical workflows.
• An initial conformance test suite for independent verification.
• Architectural guidelines ensuring the protocol remains inspectable and protected from commercial capture.

Homepage

Homepage/Blog

Shortname

ai-computed-provenance

Participation

Leadership

Chairs

• Hans-Martin Will

Links

 Mailing List

public-ai-computed-provenance