YouTube is rolling out likeness detection tool to combat deepfakes (www.engadget.com)

YouTube is rolling out a likeness detection tool to help creators and public figures get AI-altered deepfake videos removed. The feature—initially available to YouTube Partner Program members—requires users to submit a government ID and a short selfie so YouTube can register a facial reference. The system scans uploaded videos much like Content ID scans for copyrighted audio, flagging clips where a person’s face appears to have been modified by AI; flagged matches are presented to the claimant for review and potential takedown. Note that the tool currently targets visual face manipulation only and does not detect AI‑altered voices. For the AI/ML community this is a practical, large-scale deployment of automated content policing with important technical and ethical implications. It shows platforms are adopting biometric reference matching to combat generative-miscue, but raises privacy and security trade-offs (ID/selfie storage, false positives, verification risk). Technically, the approach is limited to visual forgery detection and can be evaded by adversarial techniques or future multimodal deepfakes, so it underscores the need for better detection models, standardized forensic datasets, robust watermarking, and multimodal defenses. Operationally, the rollout highlights how research on scalable detection must integrate with legal, UX and trust considerations to be effective in the wild.