Anthropic is the latest AI company to face a lawsuit over how it trained its models. Sony Music Publishing and Warner Chappell Music filed suit against Anthropic and its co-founders in a federal court in Northern California late on Friday, August 28, alleging that the company pirated more than 20,000 songs to train its Claude chatbot.

The case arrives amid a wider wave of copyright litigation targeting AI companies, as courts and plaintiffs continue to test the line between permissible model training and unlawful copying. Last year, Anthropic was ordered to pay $1.5 billion after being sued by authors in a separate piracy case. In the new complaint, the publishers say Anthropic has continued the same pattern of allegedly pirating copyrighted material.

Why publishers are suing Anthropic

The 48-page filing describes Anthropic’s conduct as “one of the largest and most blatant ongoing thefts of intellectual property in history” and names CEO Dario Amodei and co-founder Benjamin Mann as individual defendants alongside the company.

The plaintiffs accuse the Amodei-led firm of conducting “a brazen campaign of illegally torrenting, scraping, and downloading copyrighted works” for the purpose of training Claude.

The complaint offers further detail on the alleged acquisition of the material. Co-founder Benjamin Mann is accused of obtaining more than five million pirated books using BitTorrent, while employees allegedly downloaded another 2 million from Pirate Library Mirror. Those books reportedly contained lyrics and sheet music.

Anthropic also allegedly obtained lyrics from MusixMatch and LyricFind, both of which pay labels for licenses to their content.

The catalog Anthropic allegedly pirated

The publishers say the allegedly pirated catalog includes well-known songs such as Mariah Carey’s “All I Want for Christmas Is You,” Taylor Swift’s “Paper Rings,” Marvin Gaye and Tammi Terrell’s “Ain’t No Mountain High Enough,” Bon Jovi’s “Livin’ on a Prayer,” Earth, Wind & Fire’s “September,” and Leonard Cohen’s “Hallelujah.”

Sony and Warner are seeking a jury trial and statutory damages of $150,000 per pirated copy, plus $25,000 for each instance in which copyright information was copied from a work. The companies say that could amount to billions of dollars for more than 20,000 songs.

In cases involving individually copyrighted commercial recordings, plaintiffs do not need to prove financial loss before seeking statutory damages, which is one reason the alleged source and scale of the copying matters so much in disputes like this.

Why the piracy claim matters

The way Anthropic allegedly obtained the music catalog is central to the case. The dispute traces back to a 2025 authors’ case, in which a judge ruled that using copyrighted works to train AI can qualify as fair use, but downloading those works from pirated sources is illegal.

That ruling contributed to the $1.5 billion settlement, and Sony and Warner cite it in their complaint, arguing that Anthropic’s business model is “built on copyright theft.”

The case also underscores an irony that has become familiar in the AI industry. AI labs have strongly defended their own systems against distillation, the practice of training a rival model on another model’s outputs, calling it the theft of proprietary work. The same companies now stand accused of taking the proprietary work of others.

Anthropic’s response

Anthropic said it disagrees with the publishers’ claims and intends to defend itself in court.

“We disagree with the publishers’ claims, and we intend to defend ourselves robustly in court,” the company said. It also described the case as “the third lawsuit from the same lawyers, recycling allegations from cases already before the courts,” while reiterating that model training is “a transformative fair use.”

The lawsuit follows other cases against Anthropic filed by Universal Music Group, Concord, BMG, and Round Hill Music. As those disputes continue, the broader question of whether US courts will treat AI training on copyrighted material as fair use remains unresolved.