Provenance Records for Human–AI Work: Timestamped Hashes and Versioned Definitions as Verification Practice (Conceptual Note)

When people and AI systems produce work together, every result raises the same question: who made what, when, and has it changed since? This conceptual note describes a practice that combines three independently checkable records: a persistent identifier with a registration date, a cryptographic hash of every file, and a blockchain-anchored timestamp. Definitions are published as versions under one concept identifier so that change stays visible. The note states what such records prove and what they do not prove, and shows how the same structure can document human review of AI output: which person checked which result at which time. It is a description of a working practice, not an empirical study.