YouTube rolls out AI likeness detection for creators to combat deepfakes

YouTube has introduced an innovative likeness-detection feature aimed at safeguarding creators within its Partner Program. This initiative follows a successful testing phase and is designed to help eligible users identify and request the removal of AI-generated videos that replicate their faces or voices without permission. A representative from YouTube confirmed that this marks the beginning of a global rollout, with creators receiving email notifications about the new tool on Tuesday morning. The primary goal of this technology is to protect creators from the unauthorized use of their likeness, which could lead to false endorsements, misinformation, or damage to their reputations. This development comes in response to alarming instances of unauthorized AI usage, such as the case involving electronics firm Elecrow, which used a deepfake of YouTuber Jeff Geerling’s voice to market its products. YouTube’s Creator Insider channel provided details on the operation of the tool. Creators can access the new “Likeness” tab, consent to data processing terms, and scan a QR code with their smartphones. This code directs them to an identity verification page where they must submit a photo ID and a brief selfie video. Upon successful verification, creators will have access to a dashboard showcasing all videos that contain their likeness. From this platform, they can submit removal requests in line with YouTube’s privacy policies, lodge copyright complaints, or archive the videos for future reference. Participation in the program remains voluntary; creators who choose not to participate will see their data scanning cease within 24 hours. Earlier in 2023, YouTube began evaluating this technology in collaboration with Creative Artists Agency (CAA), aiming to assist celebrities, athletes, and influencers in detecting deepfake content on the platform. Additionally, in April, YouTube expressed its support for the proposed NO FAKES Act, legislation intended to restrict the creation and distribution of misleading AI-generated replicas that impersonate individuals for harmful purposes.