Sony Music has asked digital platforms to remove more than 260,000 AI deepfake songs. According to the Financial Times on October 5, 2026, by the end of September Sony's takedown requests targeting generative AI tracks that imitate its artists' voices and likenesses had passed 260,000, up from 135,000 at the end of March — nearly double in six months. Artists impersonated include Harry Styles, Adele, Britney Spears, Queen, and Michael Jackson.
Sony's stance: it will keep fighting, but the fight is getting harder
Sony's response to the Financial Times has two layers. It will keep enforcing its artists' rights and expects platforms to act quickly and stop the same content from reappearing. It also concedes the strain, describing the effort as becoming an uphill struggle. Sony characterizes these works as generative AI deepfake tracks that imitate its artists' voices and likenesses without permission, says the problem is happening at scale, and argues it harms artists, misleads fans, and undermines the ethical use of AI.
The figure needs careful reading: 260,000 is the number of requests Sony has sent, not the number of tracks platforms have removed. The Financial Times did not report how many were taken down, how fast services acted, or which platforms received the requests. The request count measures how much infringing content is being produced; only removal numbers would measure whether enforcement is working.
Who pockets the money, and why takedowns cannot keep up
The harm is not only deception. A Sony spokesperson told Billboard that every fraudulent stream takes money away from the artists and songwriters whose music is being listened to. Streaming pays by the play, so fake tracks that slip onto real artists' pages, playlists, and recommendation feeds divert real revenue, while fans may finish a counterfeit song without ever knowing.
Production, meanwhile, costs almost nothing: generate tracks in bulk with AI music tools, dress them up with similar artist names and artwork, and upload. A fake song does not need to be a hit; sitting in a catalog and quietly collecting plays is enough. Removal is the opposite kind of work, entirely manual: find each track, document it, file a request, wait for the platform, then watch for the same song returning under a new name. Music Business Worldwide calculates that Sony averaged more than 20,000 takedown requests a month from April through September. This is no longer an occasional legal task; it is a permanent production line.
The same problem, attacked from the other end
Sony's route is after-the-fact removal: broad, but slow and expensive. The other route is securing rights in advance, turning a voice into a protected asset before it is copied. We recently covered how Meloni moved to trademark her own voice with the EU, with Italy's prime minister using a trademark filing as a legal shield. Put side by side, both stories describe the same shift: when imitating a voice costs nearly nothing, protection can no longer rely on handling one case at a time.
On the legislative side, the proposed US NO FAKES Act, which would give individuals control over digital replicas of their voice and likeness, has been introduced repeatedly and remains in progress. Until stronger rules land, upload filters, rights registries, and label takedown pipelines will keep racing against generation speed — and the 260,000 figure will very likely be higher next time it is updated.
