YouTube Fights Deepfakes With New AI Detection Tool

YouTube is enhancing its efforts against AI-generated deepfakes by providing a specialized detection tool to celebrities. This initiative aims to curb the increasing trend of unauthorized deepfakes that leverage famous individuals' likenesses. The platform is granting access to this tool, which has been in development for two years, to celebrities at high risk.

The detection technology operates on a principle similar to YouTube's Content ID, enabling it to scan content and flag potentially infringing material. Celebrities or their representatives can opt into the program by uploading their likeness, allowing the tool to identify AI-generated videos that mimic them. This measure seeks to offer a foundational layer of protection for individuals whose identities are targeted by AI.

While the tool can flag problematic content, YouTube's Chief Business Officer, Mary Ellen Coe, noted that not all flagged videos will be removed. Exceptions are made for content like parody and satire, in line with community guidelines. However, content that directly replaces an individual's likeness and could harm their livelihood is subject to takedown.

This expanded rollout follows a pilot program that began last year with major creators and was later extended to politicians. YouTube is now making this tool broadly available to actors, athletes, musicians, and other public figures. Executives indicated that during the pilot, a small fraction of flagged content, primarily negative or disparaging media, was removed by creators.