YouTube expands its AI likeness detection technology to celebrities

YouTube has announced an expansion of its innovative "likeness detection" technology, designed to identify AI-generated content, including deepfakes, that involves individuals in the entertainment sector. This advancement, revealed on Tuesday, mirrors the platform's existing Content ID system, which effectively detects copyright-protected material in user-uploaded videos. With likeness detection, the focus shifts to safeguarding creators and public figures from unauthorized use of their identities—a growing concern for celebrities often targeted by scam advertisements. Initially introduced to a limited group of YouTube creators in a pilot program last year, the technology has since broadened its reach. It now includes politicians, government officials, and journalists, with the latest announcement extending its availability to talent agencies, management firms, and their celebrity clients. Major agencies like CAA, UTA, WME, and Untitled Management have actively provided feedback on the new tool, aiding its development. Importantly, utilizing the likeness detection feature does not necessitate that entertainers maintain their own YouTube channels. Instead, the tool scans for AI-generated content to identify visual matches with the faces of enrolled participants. If a match is detected, users have the option to request the removal of the video for privacy policy violations, submit a copyright removal request, or take no action at all. However, YouTube has clarified that not all content will be removed, as it allows for parody and satire under its guidelines. Looking ahead, YouTube plans to enhance this technology further by incorporating audio detection capabilities as well. In tandem with these advancements, the company is advocating for similar protections at the federal level, supporting the NO FAKES Act in Washington D.C., which seeks to regulate the unauthorized use of AI to recreate an individual's voice and likeness. Although YouTube has not disclosed how many AI deepfake removals have been executed via this tool, it acknowledged in March that the figures remain relatively low.