YouTube's new facial recognition tool aims to prevent AI deepfakes

Content creators can search for AI deepfakes of themselves on YouTube and have such videos removed. YouTube requires identification and a facial video for this.

listen Print view
Man winks for selfie

(Image: Roger Lew CC-BY 2.0)

3 min. read

YouTube apparently wants to take stronger action against the spread of videos with people created or altered by artificial intelligence (AI). To this end, the platform is now offering video creators a similarity detection tool. This allows content creators to search for videos on YouTube that contain their own face, for example via AI deepfake. The user can then decide whether to request the removal of this video or accept it.

Before YouTube can detect similarities in videos, the service requires the upload or photograph of an ID or driver's license with a photo, as well as a short video of the person's face from multiple perspectives. The portal wants to verify identity with this. However, verification can take up to five days. Afterwards, YouTube Studio lists found videos that contain one's own face or similarities, which can then be checked by the user.

This tool against deepfakes has been under development for over a year. In early September 2024, YouTube announced that more protection against deepfakes and unauthorized voice cloning is planned. The similarity detection system will be rolled out in the coming weeks, but YouTube still describes it as an "experimental feature" to gather user feedback. The similarity detection tool will also only be available in selected countries at least initially, but is expected to be expanded to other regions.

Videos by heise

The similarity detection works according to YouTube similarly to the launched in 2018 as a measure against online piracy tool "Content ID". With this system, YouTubers can more easily detect and manage their copyrighted content on the video portal. At the request of the copyright holders, the content to be protected is stored in a database with audio and image files and compared with newly uploaded videos.

YouTube also emphasizes the difference between privacy protection and copyright infringement. In both cases, content creators can request the removal of foreign videos, but for different reasons. Examples of copyright infringement cited by YouTube include copying one's own videos into foreign videos or video clips in foreign videos where the audio track has been changed to give the impression that one is saying something else.

Empfohlener redaktioneller Inhalt

Mit Ihrer Zustimmung wird hier ein externes YouTube-Video (Google Ireland Limited) geladen.

Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (Google Ireland Limited) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.

Conversely, privacy violations occur when someone else produces or copies one's own face, either via AI or by copying it, and puts foreign words into the mouth of the copied person, for example to represent political opinions or to promote products. These cases are now to be recognized and thus prevented more quickly by the similarity detection tool.

(fds)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.