TikTok unveils its anti-deepfake technology—the timing is no coincidence.

AI-generated clones are multiplying faster than the tools supposed to contain them. TikTok confirms that it is testing a likeness detection feature with a small group of American creators on a voluntary basis: the system identifies videos that reuse their face without authorization.The approach follows YouTube’s, and the European timetable provides an explanation at least as convincing as good intentions.

A clone detector reserved for a handful of American creators

The opt-in creator begins by proving their identity through Jumio, a specialist provider, by combining a real-time selfie and an ID document. The platform then scans AI-generated content likely to reproduce their features and presents the matches to them. They review the list and report videos or accounts that impersonate them. A sort of Content ID for faces: just as YouTube recognizes protected music to manage the associated rights, TikTok promises to recognize a face to block impersonations.

TikTok says it does not retain identity documents and uses facial data only to identify unauthorized content. The statement is worth keeping in mind, because the system relies on an acknowledged paradox: entrusting your face and identity papers to a platform to protect yourself from copies of your face. The climate helps explain the urgency: in early July, Meta launched Muse Image, a generator that by default draws on public Instagram photos, with accounts automatically enrolled. The backlash was immediate. TikTok’s test, for its part, remains limited to the United States, with no timetable whatsoever for Europe.

On August 2, Europe begins handing out fines

YouTube paved the way: its equivalent tool, long confined to a limited circle, was expanded to major creators, then to celebrities and talent agencies. TikTok followed suit at the precise moment when the European framework tightened. Article 50 of the AI Act applies from August 2: all deepfakes must be labeled as artificially generated or manipulated, and synthetic content must bear machine-readable markings. Brussels has granted a grace period until December 2, after which fines may climb to €15 million or 3% of worldwide turnover. For a platform the size of TikTok, the second option hurts considerably more.

A likeness detector does not, by itself, tick all these boxes: the text requires labeling and information for internet users, not merely a reporting button. The tool will, however, bolster the record of good conduct that each platform must present to European regulators.Providing protection for cloned people before the deadline, rather than after a formal notice, is simply sound prudence.

French creators will have to wait their turn, with one certainty serving as a consolation prize: if the tool ever crosses the Atlantic, it will be less out of generosity than under Brussels’ watchful eye.

👉🏻 Follow tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.

Opera One - AI-powered web browser

By: Opera

Download for free

Source: Matt Navarra/X