YouTube’s AI ‘Likeness Detection’ tool searches for deepfakes of popular creators

Starting today, YouTube Partner Program creators have access to a new AI detection feature that will allow them to find and report unauthorized uploads using their image. As this YouTube video shows, after verifying their identity, creators can view flagged videos in the Content Detection tab of YouTube Studio. If a video resembles unauthorized AI-generated content, creators can submit a removal request.

The first wave of eligible creators were notified via email this morning, and the feature will roll out to more creators over the coming months. YouTube warned early adopters in a guide to the feature that, in its current state of development, it “can show videos featuring your real face, not edited or synthetic versions,” such as clips of a creator’s own content. It works similar to Content ID, which YouTube uses to detect copyrighted audio and video content.

YouTube initially announced the feature last year and began testing it in December in a pilot program with talent represented by Creative Artists Agency (CAA). YouTube’s blog post at the time noted: “Through this collaboration, many of the world’s most influential people will have access to cutting-edge technology designed to identify and manage AI-generated content that features their likeness, including their face, on YouTube at scale.” »

YouTube and Google are among many tech companies offering AI video generation and editing tools, and the resemblance detection tool isn’t their only feature in development to handle AI-generated content on the platform. Last March, YouTube also began requiring creators to label uploads that included AI-generated or edited content and announced a strict policy regarding AI-generated music “that imitates an artist’s unique singing or rapping voice.”