Sony Music Publishing, Warner Chappell, and a coalition of other music publishers dropped a lawsuit on Anthropic late Friday, accusing the AI lab of running what they call a “brazen campaign of illegally torrenting, scraping, and downloading copyrighted works.” Filed in the Northern District of California, the suit names co-founders Dario Amodei and Benjamin Mann as defendants alongside the company. The publishers don’t mince words — they’re calling it “blatant theft” and “flagrant piracy,” alleging Anthropic grabbed millions of book copies containing lyrics and sheet music to train Claude.
Anthropic isn’t backing down. “We disagree with the publishers’ claims and we intend to defend ourselves robustly in court,” a spokesperson told TechCrunch. That’s the standard defiant posture, but the stakes here are higher than typical AI copyright bickering. Some of the same lawyers behind this suit also represented Concord Music Group and Universal Music Group in a January case, and they led the Bartz v. Anthropic litigation that ended with a $1.5 billion judgment against the company. The Bartz ruling was nuanced: a judge said using copyrighted works for training was legal, but acquiring that content through piracy was not.
That distinction is exactly what makes this new case dangerous for Anthropic. The publishers aren’t just arguing about fair use or training data — they’re alleging torrenting, which is a fundamentally different legal beast. If the court finds Anthropic actually downloaded pirated copies at scale, the Bartz precedent gives plaintiffs a clear path to enormous damages. The lawsuit is also notably broad, covering lyrics and sheet music embedded in books rather than just audio recordings, which expands the potential damages pool significantly.
For Anthropic, the timing couldn’t be worse. The company is trying to position itself as the responsible, safety-focused alternative in the AI race while simultaneously fighting multi-front legal battles over how it built its models. The Bartz settlement already cost them $1.5 billion. Another major loss — especially one involving major music publishers with deep pockets and aggressive legal teams — could force changes to how Anthropic sources training data altogether. Whether that means licensing deals with publishers or more carefully vetted datasets remains an open question, but the era of scraping first and apologizing later is clearly ending.
