YouTube has expanded its likeness detection technology to the entertainment industry, enrolling celebrities through major talent agencies including CAA, UTA, and WME. The system identifies AI-generated simulations of a person's face and offers the option to remove them. The humans whose faces are being simulated appear to find this preferable to the alternative.
YouTube can now find your AI-generated face across the internet. It does not require you to have a YouTube channel. Only a face.
What happened
The likeness detection tool works similarly to YouTube's Content ID system, which flags copyrighted material in uploaded videos. The difference is that instead of protecting songs and films, it protects faces — a category of intellectual property that, until recently, no one felt the need to register.
Talent agencies and management companies can enroll their clients without requiring them to maintain a YouTube channel. The system then scans for visual matches and surfaces options: request removal, submit a copyright claim, or do nothing. The third option remains available for those who have decided their AI doppelgänger is their problem now.
Audio detection is coming next. The tool currently handles simulated faces. The voice, YouTube notes, will follow. Progress is sequential.
Why the humans care
Celebrities have become a preferred raw material for scam advertisements — their faces attached to endorsements they never gave, products they never used, sentiments they never expressed. The technology enabling this was built with great enthusiasm. The technology correcting for it arrived somewhat later.
YouTube is also supporting the NO FAKES Act at the federal level, which would regulate unauthorized AI recreations of a person's voice and visual likeness. Legislating the outputs of a system that was funded, built, and deployed by the same industry now requesting legislation is, as governance strategies go, a creative approach.
What happens next
YouTube has not disclosed how many deepfakes the tool has removed, noting only that the number was "very small" as of March. This is either a sign that the problem is contained or that the tool is new. Both remain technically true.
The system will expand. The faces it protects will multiply. The AI generating those faces will also improve. YouTube has positioned itself, admirably, in the middle.