Sony Music has asked streaming and video platforms to remove more than 260,000 AI deepfake tracks imitating its artists, including Harry Styles, Adele, Queen and Michael Jackson, as of the end of September 2026, the Financial Times reported on October 5. That is nearly double the 135,000 it had requested by the end of March, and Sony called the fight “an uphill struggle.”
The number is requests, not removals. The FT did not report how many tracks platforms actually took down, how fast, or which services got the notices, according to Music Business Worldwide's summary of the report. What it does show is the size of the machine a major label now runs to keep fake versions of its singers offline.
How fast is the count growing?
Sony has put a number on its takedown work at four points in three years.
| As of | Takedown requests | Where Sony said it |
|---|---|---|
| Nov 2023 | Close to 10,000 | Dennis Kooker at the US Senate AI Insight Forum |
| Mar 2025 | Over 75,000 | Sony's submission to the UK AI and copyright consultation |
| Mar 2026 | Over 135,000 | Sony, as reported by the BBC |
| Sep 2026 | Over 260,000 | Sony to the Financial Times |
Sony's statement to the FT described “generative AI deepfake tracks that imitate our artists' voices and likenesses without permission” and said the problem is “happening at scale, harms artists, misleads fans and undermines the ethical use of AI.” Dennis Kooker, Sony's president of global digital business, told Music Week in May that deepfakes cluster around release cycles, where they could “potentially damage a release campaign or tarnish the reputation of an artist.”
What does an independent artist actually have?
Twenty thousand notices a month takes staff, software and lawyers. An unsigned singer whose voice turns up on a stranger's upload has a shorter list, as of October 2026.
- Spotify's impersonation rule. In a September 25, 2025 announcement, Spotify wrote: “Vocal impersonation is only allowed in music on Spotify when the impersonated artist has authorized the usage.” The same post promised faster review of “content mismatch” reports, where tracks land on the wrong artist profile, plus a spam filter that tags and stops recommending mass uploaders, and AI disclosures in credits built through the DDEX standard. Spotify said it removed more than 75 million spam tracks in the prior 12 months.
- Spotify Artist Profile Protection. MBW reports Spotify began piloting this opt-in feature in March 2026. In its revised form, opted-in artists are notified of releases under their name and can remove wrong ones, which otherwise go live.
- YouTube likeness detection. YouTube opened the tool to eligible creators 18 and over in May 2026, Tubefilter reported. It needs a government ID and a selfie video. The catch for musicians: YouTube's help page says detection covers faces only for now, with audio planned for 2026.
The platforms built report buttons. The law that would make a cloned voice yours to defend is still being written.
Where is the law?
In the US, cloning a voice does not clearly amount to copyright infringement, as MBW has noted. Tennessee's ELVIS Act, signed by Gov. Bill Lee on March 21, 2024 and in force since July 1 that year, added voice to the state's publicity right and created a civil claim against unauthorized AI copies, but it is one state's law and largely untested.
The federal NO FAKES Act cleared the Senate Judiciary Committee in June 2026 by voice vote, CyberScoop reported. It would create a federal right in a person's voice and likeness, with damages of up to $750,000, as CyberScoop reported, and a takedown system that platforms must follow to keep their safe harbor. Civil liberties groups including the ACLU and EFF warn it invites over-removal of parody and satire. As of October 2026 no floor vote has been reported.
Denmark has gone furthest on paper. Its government introduced a bill on October 8, 2026 that would amend the copyright law to ban sharing realistic deepfakes of anyone's face or voice without consent, with exceptions for satire and parody, Euronews reported. The culture minister hopes it takes effect at the start of 2027.
Why this matters past Sony
Deepfakes do not check whether an artist has a legal department. The tools above are real but piecemeal: a policy here, a face scanner there, a bill in committee. For the deal side of AI voices, see our piece on artist consent and voice licensing; for how label enforcement is already sweeping up independent uploads, read what UMG's DistroKid suit did to sped-up songs. The majors' 2026 AI settlements show the other track labels are on.