"Only I have the right to be me." It is with this statement, attributed to Neal Mohan, YouTube's CEO, that the Alphabet subsidiary announced on Thursday evening, April 23, 2026 – Friday morning for Europe – the expansion of its artificial intelligence deepfake detection tool, named "Likeness Detection," to Hollywood. Previously reserved for a circle of about 200 partner creators (mainly YouTubers with over 10 million subscribers and some athletes), the technology is now freely available to any celebrity, actor, musician, athlete, or public figure who can prove their identity through an enhanced verification process.

A mechanism inspired by Content ID

Technically, Likeness Detection is directly inspired by Content ID, the tool implemented by YouTube as early as 2007 to identify violations of music and audiovisual rights. The principle: the owner of a face or voice submits several reference samples (videos, photos, audio files), from which the system trains a proprietary biometric fingerprint. Algorithms then continuously scan new uploads – approximately 720,000 hours of video per day worldwide – and automatically flag content in which the person's voice or face appears in a form suspected of being manipulated by AI. The celebrity then receives a notification and can request the removal of the content, monetization in their favor, or explicit authorization – each option being legally tracked.

The stake: exploding deepfakes

The context is striking. According to the Sensity AI barometer published in March 2026, the number of deepfakes online featuring celebrities has surged by 530% between 2024 and 2025, and is expected to rise by 700% this year. Nearly 96% of these videos are non-consensual pornography, primarily targeting women (Taylor Swift, Scarlett Johansson, Jenna Ortega have all been victims of massive campaigns in recent months). The remainder is divided between scams (fake Elon Musk, fake Brad Pitt promising financial returns), political propaganda (fake Joe Biden, fake Donald Trump broadcast during campaigns), and satire. The American union SAG-AFTRA, which made image rights its main battle during the 2023 strike, hailed it as "a major but insufficient step forward."

Towards federal legislation?

YouTube's announcement comes as the US Congress is in the final reading of the "Take It Down Act," a bipartisan bill sponsored by Ted Cruz (Republican) and Amy Klobuchar (Democrat), which would require all platforms to remove any non-consensual intimate content – including deepfakes – within 48 hours. The bill, already adopted by the Senate with 97 votes to 3, is expected in the House in the coming weeks. In parallel, the "TVPA Act" (Taking Voice and Persona to Aid) specifically aims to extend federal rights to voice and image. For YouTube, anticipating these regulations is also a commercial ahead of the curve: Meta, TikTok, and X will have to follow suit or risk massive fines from the Federal Trade Commission.

A caveat: not everything will be removed

However, YouTube is careful to specify in its statement published on the official blog that "not all flagged content will be automatically removed." Content deemed to fall under political satire, cultural criticism, or education may be kept online, in accordance with community guidelines. The platform therefore reserves a discretionary power that already worries victim advocacy groups. "This is good news, but the devil will be in the details of implementation," tempers Suzanne Verdon, a specialized lawyer at the EFF. The first reports in the coming weeks will show whether the balance tips more towards the effective protection of victims or the preservation of potentially abusive editorial freedom.

Editorial Opinion

The expansion of Likeness Detection to Hollywood is good news, but it also reveals the scale of the delay by platforms. For three years, the deepfake market has exploded without any major player investing seriously in large-scale detection – due to technical cowardice, economic calculation, and sometimes contempt for the most vulnerable victims. YouTube, by relying on the proven infrastructure of Content ID, demonstrates that it was perfectly possible to act sooner. The real question now is elsewhere: what about the tens of millions of people who are not famous, who do not have a SAG-AFTRA stamp or a legal team, and who are already victims of intimate deepfakes today? The technology now exists; there must be a legal obligation for universal, free, and accessible access. Without it, YouTube's tool will remain what it is: a privilege for the powerful, when it should be a right for all.

Key takeaways

  • YouTube opens its "Likeness Detection" tool to all Hollywood celebrities, free of charge, on April 24, 2026.
  • Technology inspired by Content ID: biometric fingerprint of face and voice.
  • Rise in deepfakes featuring celebrities: +530% between 2024 and 2025 (Sensity AI).
  • 96% of deepfakes are non-consensual pornography, primarily targeting women.
  • Not all flagged content will be removed: political satire and cultural criticism preserved.