The system requires participants to submit an ID and a selfie video of themselves. (Likeness detection is focused on faces specifically, as opposed to a voice or other identifying characteristics.) Removal of deepfakes isn’t guaranteed, and there are protected use cases like parody or satire. YouTube has previously said that when content creators used the feature, they requested only a “very small” number of videos of themselves be removed.

YouTube has compared likeness detection to Content ID, its system for finding (and removing) copyrighted material on the platform. The difference is that with Content ID, rights holders can opt to monetize other users’ videos that use their material and split the revenue. That’s not yet possible with likeness detection, but that clearly seems like the direction the industry is moving toward.

Earlier this month, YouTube announced a feature allowing creators to digitally clone their likeness using AI, which could then be inserted into videos. Talent agency CAA (which YouTube says supported the likeness detection expansion) has a database filled with clients’ biometric data that entertainers can retain — or deploy for commercial opportunities. TikTok star Khaby Lame effectively sold off the rights to his likeness, which would then be used to sell products online. (The deal has run into several road bumps and it’s not clear if it has closed, according to Business Insider.)