YouTube has started rolling out a new tool designed to detect face manipulation in videos. This feature is aimed at combating unwanted deepfake materials and facilitating their removal from the platform. Engadget reports on this development.
Currently, access to this tool is limited to members of the YouTube Partner Program. The system targets cases where artificial intelligence alters a person's face. However, it does not yet detect voice alterations made without the consent of the individual.
To utilize this protective feature, users must provide YouTube with identity verification: upload a copy of a government document and a brief video selfie. They will also need to supply the system with original materials for verification.
Once submitted, the system operates on a principle similar to Content ID, which is used to identify copyrighted audio. It scans uploaded videos for potential matches and allows face owners to review results and mark violations for removal.
The launch of this tool comes amid growing concerns regarding the proliferation of video generation technologies, especially after the emergence of new models like Sora 2 from OpenAI.