Engram Provenance: Substrate-Rooted Provenance for Safety-Critical AI Systems

This paper introduces Engram Provenance, a substrate-rooted provenance model for AI systems. Building on the Engram Layer and the Engram Signature, provenance is defined as continuity of execution-realized identity across time, perturbation, and substrate variation. Engram Provenance formalizes lineage as a sequence of stable Engram Signatures and provides minimal conditions for auditability, chain-of-custody, and substrate-rooted trust in safety-critical domains