Sony's AI Deepfake Takedowns Top 260,000 Tracks

Sony Music says its AI deepfake takedowns have blown past a quarter of a million, nearly double the total it reported only six months ago. The number landed this week after the label shared it with the Financial Times, and it paints a picture of an industry stuck in a endless game of whack-a-mole.

If you stream music, you have probably already scrolled past a fake. A track that claims to be your favorite artist, a cover that sounds weirdly right, a song that never appeared on any official release. Those fakes are exactly what Sony is trying to scrub from the internet, and the pace is accelerating fast.

The numbers keep climbing

According to Music Ally, Sony said in March of last year that it had gotten more than about seventy-five thousand AI-generated deepfakes of its artists pulled from online platforms. By this past March the count of takedown notices had reached more than about one hundred thirty-five thousand. Now it is past a quarter of a million, per the label's latest update.

That is almost a doubling in half a year. Music Ally notes that the label has a clear motive for sharing the stat: it wants more support from policymakers as it fights the problem. In its own words to the FT, Sony said it will keep enforcing its artists' rights and expects platforms to act quickly and stop the same content from reappearing, but called it "an uphill struggle."

Why the fakes keep coming back

The core issue is repetition. A track can be removed from one service and pop up again somewhere else, sometimes within hours. As RouteNote reports, removed tracks can reappear elsewhere, which makes the fight about prevention as much as deletion.

The scale of uploads makes it worse. RouteNote points out that in July, Deezer reported detecting around about ninety thousand AI-generated tracks arriving on its platform every single day. Not all of those are fraud or impersonation, but the flood makes spotting the bad ones far harder for platforms, distributors and rights holders.

It is not just about songs

Deepfakes have spread well beyond audio. Music Ally recalls country star Martina McBride saying a fan nearly lost his home to a scammer using AI to pretend to be her. TikTok also dealt with a wave of deepfaked ads using stars like Taylor Swift and Rihanna as bait for phishing scams.

Swift also had to fend off a completely fake music video and song released in March on a YouTube channel built around that kind of content. The channel has since been shut down. For artists, the damage is not only lost streams, it is fans being tricked in her name.

What artists and lawmakers are doing

In the US, the No Fakes Act, which targets non-consensual deepfakes, ran out of legislative time before the election cycle two years back. It has since been reintroduced, and Music Ally reported in June that it took another step toward becoming law. Nothing is final yet.

Meanwhile, artists are taking matters into their own hands. Swift filed trademark applications to protect her voice and image in April, followed by Lionel Richie and the Backstreet Boys in June. It is a stopgap while the legal system catches up.

Platforms are also adding guardrails. RouteNote says Deezer has introduced systems to identify and label AI-generated music, while Apple Music has rolled out AI transparency tags. Independent artists without a major label's legal team are advised to watch their profiles, document suspicious listings and report them to the platform or distributor.

What this means for your playlists

Expect more labels on streaming apps and more pressure on services to block repeat uploads. Sony's own admission that the moles are multiplying shows takedowns alone will not fix this. Until platforms can stop fakes at the door, the safest habit is to check who actually released that surprise track before you hit share. For more on how AI is reshaping the industry, see our music coverage and our AI news hub.

The story first surfaced in the Financial Times, according to Music Ally, and RouteNote followed with its own breakdown. Both outlets stress the same point: the volume of AI-generated fakes is growing faster than anyone can remove them.

Think about what that means for a working musician. Every fake upload steals a little attention, a few streams and some trust from the real artist. Multiply that by hundreds of thousands of notices and the takedown tally starts to look like a measure of how big the impersonation economy has become.

There is also a fairness question. A giant label like Sony can staff teams to hunt fakes, but a bedroom producer cannot. That gap is why distributors and platforms, not individual artists, are being pushed to build detection into the upload process itself. Stronger tools could protect smaller creators who have the most to lose.

Fans have a role too. Reporting a suspicious track takes seconds, and platforms rely on those flags to catch fakes that slip past automated filters. If a song feels off, trust that instinct, check the artist's official channels and report it. Small habits like that add up when the volume is this high.