An Open Standard for AI Licensing Provenance and Agent-to-Agent Trust
AIMS is a focused standard for AI training data licensing provenance, runtime content access rights, and agent-to-agent cryptographic trust. While Model Cards document performance and Dataset Cards document composition, AIMS answers the licensing questions they don't address.
Training Data Licensing Provenance
What's the licensing status of the training data? Can it be proven?
Runtime Content Access Rights
What content can the system legally access during inference?
Agent-to-Agent Trust
How do AI agents cryptographically verify each other's identity and licensing boundaries?
AIMS builds on W3C standards (DIDs, Verifiable Credentials) and works alongside Model Cards, Dataset Cards, A2A Agent Cards, and MCP for comprehensive AI transparency.
Incomplete Draft Specification: Complete for content usage transparency and training data licensing. Cryptographic components (DID method, Merkle proofs, verification protocols) require expert contribution. See what needs expert input →
Table of Contents
Specification
Status
This is an incomplete draft specification intended to invite collaboration.
Complete: Content usage transparency, training data licensing provenance, deployment context disclosure
Requires expert contribution:
- DID method specification (
did:aims) - Verification protocol bindings
- Merkle proof format for selective disclosure
- Cryptographic signature schemes
See Open Questions for the full list.
Contributing
We welcome contributions from AI developers, content publishers, standards bodies, policymakers, and civil society organizations.
To contribute:
- Open an issue to discuss proposed changes
- Submit a pull request with your contribution
- Join the discussion in existing issues
About OpenAttribution
OpenAttribution is a coalition of publishers, brands, and technology providers developing standards for content licensing and AI transparency.