Sony and Warner Chappell Sue Anthropic Over Song Lyrics in Claude
Two of the three biggest music publishers filed a broad copyright suit against Anthropic and two of its founders, focusing on how the training material was obtained rather than on whether training itself is legal.
Sony Music Publishing and Warner Chappell Music, along with dozens of affiliated publishers, sued Anthropic late on Friday in the US District Court for the Northern District of California. The complaint also names CEO Dario Amodei and co-founder Benjamin Mann personally, and accuses the company of a “brazen campaign of illegally torrenting, scraping, and downloading copyrighted works” to train Claude. Anthropic told TechCrunch it disagrees with the claims and intends to defend itself robustly in court.
The filing runs to 48 pages and lists 24 Sony affiliated companies and 11 Warner Chappell affiliated ones as plaintiffs. It alleges that Anthropic acquired and reproduced tens of thousands of musical works without permission, that it stripped out copyright management information, meaning the credit and ownership data attached to a file, and that Claude reproduced lyrics as output. Named songs include “Eye of the Tiger,” “September” and Taylor Swift’s “Paper Rings.” The publishers are seeking statutory damages of up to 150,000 dollars per work for wilful infringement, plus up to 25,000 dollars for each removal of copyright management information. Anthropic has been here before. Universal, Concord and ABKCO sued over lyrics in October 2023 and again in January 2026, BMG filed in March, and Round Hill Music filed on 17 August. Some of the lawyers on this case also ran Bartz v. Anthropic, the authors’ case that ended in an approved 1.5 billion dollar settlement in July.
What’s actually going on here: the Bartz outcome is the key to reading this one. In that case a judge drew a line that has shaped every filing since: using copyrighted works to train a model can be lawful, but obtaining them through piracy is not. So the fight has largely moved off the philosophical question of whether training is fair use and onto a much more concrete one, namely where the files came from. That is why this complaint leans so hard on torrenting and on shadow libraries rather than on the model’s behaviour, and it is also why the founders are named individually. Both details make it a procurement case as much as a copyright case. For a company that has settled once at 1.5 billion dollars, the arithmetic of “per work” damages across tens of thousands of compositions is the real exposure here.
What this means for you: nothing changes about using Claude today, and there is no reason to expect an interruption. What is worth watching is the direction of travel. If courts keep separating “you may learn from it” from “you may not steal the copy you learned from,” the practical result is that AI labs will need paid, documented data deals, and that cost eventually shows up in subscription prices. For anyone who makes things for a living, whether that is music, writing or photography, this is the line of cases actually deciding what happens to your work, more than any of the louder debates about creativity and machines. And if you build with these tools professionally, the lesson is the same one every large model provider is now learning in public: keep a record of where your material came from.
Sources
Someone Finally Asked Workers Directly: 3% Say AI Cost Them a Job
A sociologist commissioned YouGov to ask 1,250 US workers whether AI had changed their job since 2023. About 3% had lost one, 6% got a job that did not exist before, and 9% got an AI related promotion.
