Celebrities will be able to find and request removal of AI deepfakes on YouTube

At a glance:

• YouTube expands its likeness‑detection tool to cover celebrities regardless of whether they have a channel.
• Enrolled public figures can search for AI deepfakes of themselves and request takedowns, though removal is not guaranteed.
• Participation requires a government‑issued ID and a selfie video; protected uses such as parody remain exempt.

What the program does

YouTube’s likeness‑detection system scans the entire platform for AI‑generated videos that replicate a public figure’s face. When a match is found, the content is flagged for the enrolled celebrity, who can then review the video and submit a removal request. The request is evaluated against YouTube’s privacy policy, meaning not every flagged video will be taken down. The company says the feature works even for celebrities who never created a YouTube account, extending the reach of its earlier pilot that began with content creators in late 2023.

The rollout follows a March expansion that opened the tool to politicians and journalists. Early data from the creator‑only test indicated that participants asked for only a “very small” number of videos to be removed, suggesting that the majority of AI‑generated content either falls under allowed categories like satire or simply evades detection.

How it works and what participants must provide

To join the program, a public figure must submit a government‑issued identification document and a short selfie video. The system focuses exclusively on facial features; voice‑only deepfakes are not covered. Once enrolled, the likeness‑detection algorithm continuously monitors new uploads, comparing facial embeddings against the submitted reference data. When a potential match appears, the platform notifies the participant through a dashboard similar to Content ID, YouTube’s long‑standing copyright‑matching service.

Unlike Content ID, which lets rights holders monetize or block infringing videos, likeness detection currently offers only a takedown request pathway. YouTube has hinted that future iterations could allow revenue sharing or other controls, but those capabilities remain speculative at this stage.

Industry context and future implications

The move arrives as talent agencies, notably Creative Artists Agency (CAA), begin building biometric databases for their clients. CAA’s repository could enable entertainers to license their digital likenesses for commercial projects, mirroring a recent YouTube feature that lets creators clone their own faces with AI for use in videos. The same technology that powers deepfakes is being packaged as a revenue‑generating asset, a trend highlighted by TikTok star Khaby Lame’s attempt to sell the rights to his likeness for product endorsements.

Industry observers, quoted in The Hollywood Reporter, argue that the proliferation of AI‑generated celebrity content could become a new fan‑engagement channel—provided creators receive compensation. Some managers envision a future where deepfakes are tolerated or even encouraged, while others see the need for stricter controls to protect brand integrity. YouTube’s current tool is a first step toward a broader ecosystem where facial likenesses are both monetized and defended.

What remains uncertain

Removal requests are still subject to YouTube’s privacy policy, and the platform has not disclosed clear metrics on approval rates for celebrity takedowns. Protected uses such as parody, satire, or news reporting remain exempt, meaning a substantial portion of AI‑generated content may stay online even after a request. Moreover, the requirement for an ID and selfie video could deter some public figures who are wary of sharing biometric data with a tech giant.

Future updates may introduce monetization options similar to Content ID, allowing celebrities to earn from authorized uses of their AI‑generated likenesses. Until then, the tool functions primarily as a defensive measure, giving high‑profile individuals a way to monitor and, when possible, erase unauthorized deepfakes from the world’s largest video platform.