YouTube’s likeness detection has arrived to help stop AI doppelgängers (arstechnica.com)

YouTube has begun rolling out a “likeness detection” feature for creators to help flag face‑stealing AI videos. The tool, now moving beyond an initial small test group, lives in the Content detection menu in YouTube Studio and is modeled after the platform’s copyright‑matching system. Eligible creators can enroll by verifying identity—submitting a government ID and a short face video—so YouTube can build a reference for automatically detecting videos that impersonate them. The feature is currently beta and appears aimed at single‑host channels; not all creators will see it yet. This matters because realistic synthetic video is an increasing source of misinformation, harassment, and brand damage, and platforms need scalable defenses short of banning generative AI. Technically, the system looks like a biometric face‑matching pipeline that links enrolled identities to uploaded or existing content to surface suspected deepfakes. That approach raises tradeoffs: it could help creators quickly take down harmful impostor content, but it also requires handing sensitive ID and biometric data to Google, invites questions about retention, false positives, adversarial bypasses, and how well it will scale. The rollout signals YouTube’s preference for detection and mitigation tools over prohibitions, but it also kickstarts debates about privacy, security, and the arms race between synthetic‑media generation and detection.