Report: Sony Music doubles AI deepfake takedowns to more than 260000

Sony Music is stepping up its fight against streaming fraud after revealing it has nearly doubled its takedown requests for AI-generated tracks in the space of six months.

The major told the FT it had asked platforms to remove more than 260,000 tracks by the end of September – up from 135,000 at the end of March – as it contends with a surge in deepfake songs impersonating artists from Harry Styles to Adele.

Sony described the issue of "generative AI deepfake tracks that imitate our artists' voices and likenesses without permission" as "happening at scale", adding that it "harms artists, misleads fans and undermines the ethical use of AI".

We will keep enforcing our artists’ rights and expect platforms to act quickly and to prevent the same content from reappearing. But it’s becoming an uphill struggle Sony Music

We will keep enforcing our artists’ rights and expect platforms to act quickly and to prevent the same content from reappearing. But it’s becoming an uphill struggle

The label added: “We will keep enforcing our artists’ rights and expect platforms to act quickly and to prevent the same content from reappearing. But it’s becoming an uphill struggle.”

Sony, Universal and Warner were among the music companies to back the IFPI's Streaming Integrity Initiative (SII), a new industry commitment launched last month to combat streaming fraud .

The rise of generative AI has heightened the concerns streaming fraud with Deezer reporting earlier this year that is ingesting almost 75,000 AI-generated tracks per day , representing roughly 44% of the daily uploads.

Dennis Kooker , president, global digital business at Sony Music, told Music Week in May this year that the scale of the challenge across the streaming ecosystem remained uncertain.

“No one knows exactly, which is part of the problem,” said Kooker. “But could somewhere between 5% and 10% of the streams be fraudulent? Yeah, I think that’s possible. Unfortunately, the way that the ecosystem works, for those who really want to exploit it, they have a lot of incentive to create a lot of streams.”

The problem with deepfakes is they are demand-driven Dennis Kooker, Sony Music

The problem with deepfakes is they are demand-driven

Dennis Kooker, Sony Music

Speaking, at the IFPI Global Music Report event held in London in May, Kooker said that Sony Music had requested takedowns of 135,000 generative AI deepfakes of its artists’ music up to that point.

“The problem with deepfakes is they are demand-driven,” he said. “It is taking advantage of the fact that an artist is in cycle and is out there promoting their music. That is when we see the worst of the deepfakes appear... benefiting from the demand that the artist has created in the market with their fans, and ultimately detracting from what it is the artist is actually trying to accomplish with their fanbase.”

In the worst case, Kooker said this could “potentially damage a release campaign or tarnish the reputation of an artist”.

“So ‘good AI’ begins with protecting the market and the people whose work fuels it, which is why we’re investing in companies that are building solutions that ensure misuses are identified while also creating ethical AI partnerships to build new revenue streams for artists,” he added.