YouTube Confirms Altering Videos Using AI Without Notifying Creators

YouTube confirms using AI to alter Shorts on its platform without first informing creators, leaving some content producers feeling unsettled.

YouTube reveals using artificial intelligence to make video enhancements to content on its platform without first informing content creators. Confirmed to be applied to a subset of YouTube Shorts, the undisclosed changes have left some creators feeling unsettled about how their work is being used. It also raises questions about the overall implications of undisclosed algorithmic mediation.

“I was like, ‘Man, my hair looks strange,’” music educator and commentator Rick Beato, who has over five million subscribers on YouTube, told the BBC. He says he first suspected something was off when his own appearance in a video looked different. Upon closer inspection, he thought he looked as though his skin had been digitally retouched, as if he had used a filter.

Guitarist Rhett Shull reported similar discrepancies in his own Shorts. He noticed what he described as an artificial aesthetic in which his features had been over-sharpened. “If I wanted this terrible over-sharpening, I would have done it myself,” said Shull, who later posted a video about the issue, which has so far gained over 700,000 views.

On social media, other creators have been sharing evidence of unusual edits to their content going as far back as June. Users have reported skin textures appearing unnaturally smooth, fabric folds appearing exaggerated, and small details such as ears occasionally appearing distorted. While some of these alterations are subtle, they’re concerning enough to raise questions about what else YouTube might be changing without creators’ consent.

Rene Ritchie, YouTube’s head of editorial and creator liaison, stated on the former Twitter that YouTube has been testing “traditional machine learning technology” to “unblur, denoise, and improve clarity” in videos during processing. He compared this process to enhancements made by modern smartphones when users record video.

Some users have argued that Ritchie’s (and YouTube’s) terminology deliberately downplays the scope of the changes being made. YouTube has stressed a distinction between “traditional machine learning” and “generative AI,” the latter of which produces new content altogether.

Regardless, the company has thus far avoided revealing whether creators will be able to opt out of such changes eventually. The issue also raises questions about whether such unauthorized edits will negatively affect public trust in online content, beyond YouTube.