Stable Model Identity: Moving Beyond Names in AI Registries

AI model registries have traditionally relied on naming conventions and labels to identify models, but this approach fails when on-chain claims require permanent, verifiable records. By implementing cryptographic digests like SHA-256 combined with supply-chain provenance standards, the industry can create unforgeable artifact identities that prevent costly post-deployment corrections. This pattern, borrowed from software supply-chain systems like SLSA and OCI, ensures that model verification answers 'which model' definitively beyond just a name.