YouTube expands AI likeness detection to celebrities
▼ Summary
– YouTube is expanding its “likeness detection” technology to the entertainment industry to identify AI-generated deepfakes.
– The tool scans for AI-generated visual matches of enrolled participants’ faces, allowing removal requests for privacy violations.
– The technology was first piloted with creators, then expanded to politicians and journalists, and now includes talent agencies and celebrities.
– Use of the tool does not require entertainers to have their own YouTube channels, and it will support audio detection in the future.
– YouTube supports the federal NO FAKES Act to regulate unauthorized AI recreations of voice and likeness.
YouTube is broadening access to its AI likeness detection technology, extending the tool to celebrities and other entertainment industry figures, the platform confirmed on Tuesday.
The system functions much like YouTube’s existing Content ID mechanism, which scans uploaded videos for copyrighted material and allows rights holders to either block the content or claim ad revenue. Likeness detection applies a similar logic to AI-generated faces, identifying deepfakes and other synthetic representations that mimic real people. The goal is to curb the unauthorized use of someone’s identity, a persistent issue for public figures whose faces are often hijacked in scam advertisements without their consent.
Initially rolled out to a select group of creators in a pilot program last year, the feature expanded earlier this spring to include politicians, government officials, and journalists. Now, YouTube is opening it up to the broader entertainment sector, including talent agencies, management firms, and the celebrities they represent. Major players such as CAA, UTA, WME, and Untitled Management have already provided feedback on the tool’s development.
Notably, entertainers do not need to have their own YouTube channels to enroll in the program. Instead, the tool automatically scans for AI-generated visuals that match an enrolled participant’s face. When a match is found, users can choose to request removal under privacy policy violations, submit a copyright takedown notice, or take no action. YouTube stressed that not all flagged content will be removed, as the platform allows for parody and satire under its guidelines.
Looking ahead, the company plans to expand the technology to detect AI-generated audio as well. On the policy front, YouTube has also been pushing for federal protections, backing the NO FAKES Act in Washington, D. C., which aims to regulate the unauthorized use of AI to replicate an individual’s voice or likeness.
YouTube has not disclosed how many deepfake removals the tool has handled so far, but noted in March that the volume remained “very small.”
(Source: TechCrunch)
