Sony Music and Warner Chappell Sue Anthropic: The Music Industry’s New Front in the AI Copyright Wars

Estimated reading time: 5 minutes

On August 29, 2026, Sony Music Publishing and Warner Chappell Music — two of the largest music publishers in the world — filed suit against Anthropic, the maker of the Claude AI chatbot, in the U.S. District Court for the Northern District of California. The complaint, joined by co-founders Dario Amodei and Benjamin Mann as individual defendants, accuses the company of running a “brazen campaign” of illegal torrenting, scraping, and downloading of copyrighted works — including books containing song lyrics and sheet music — to train its large language models.

This is not Anthropic’s first brush with the music industry, and it is only the latest chapter in a broader legal reckoning between AI developers and rightsholders over how generative models are trained.

A Pattern of Litigation

The Sony/Warner Chappell suit builds on a case that began nearly three years earlier. In October 2023, Concord Music Group, Universal Music Publishing Group, and ABKCO Music sued Anthropic over the unauthorized reproduction of roughly 500 song lyrics by Claude. That case has since been amended twice, most recently in a 71-page filing submitted on July 22, 2026. The publishers say discovery has produced damning internal evidence: Anthropic co-founder Tom Brown allegedly queried Claude for Bob Dylan lyrics, while co-founders Ben Mann and Jared Kaplan discussed stripping copyright notices from training data. According to the complaint, the word “lyric” appeared in over 170,000 internal Claude records during a single nine-day window in September 2023. The publishers are seeking statutory damages of up to $150,000 per infringed work — and a separate, related suit filed in January 2026 covers more than 20,000 songs, with damages sought in excess of $3 billion.

Crucially, all of this arrives just weeks after Anthropic’s $1.5 billion settlement with a class of book authors (the Bartz v. Anthropic case) received final court approval on July 20, 2026. That case set an important precedent: U.S. District Judge William Alsup ruled that training an AI model on lawfully acquired copyrighted books can qualify as fair use — but downloading millions of books from pirate repositories such as Library Genesis and “Books3” is not protected, however the resulting model is later used. Authors ultimately received roughly $3,000 per pirated book, in one of the largest copyright recoveries in U.S. history.

Sony Music Publishing and Warner Chappell are now asking the courts to apply that same piracy-versus-fair-use distinction to song lyrics. Anthropic has said it disagrees with the publishers’ claims and intends to “defend ourselves robustly in court.”

Why This Case Matters

For the music industry, lyrics occupy a special place: they are short, highly memorable, and often reproduced verbatim by chatbots when users ask for them — unlike, say, a paraphrased summary of a novel. That makes “memorization” claims easier to demonstrate than in many other AI copyright disputes, and it is why music publishers have been among the most aggressive litigants against generative AI companies, alongside authors, news organizations, and visual artists.

For AI developers, the case underscores a lesson that is quickly becoming settled law: how training data was obtained matters as much as how it is used. Scraping or torrenting copyrighted works from unauthorized sources carries legal exposure that licensing simply does not.

History Repeats: Lessons from the Napster Wars

None of this is entirely new. Twenty-five years earlier, the music industry fought an eerily similar battle over a different disruptive technology: peer-to-peer file sharing.

Metallica v. Napster, Inc. (2000). In early 2000, Metallica discovered that a demo of their unreleased song “I Disappear” was circulating widely on Napster before its official release. Drummer Lars Ulrich led the band’s response, filing suit in April 2000 for copyright infringement and racketeering violations, and — in a move that generated enormous controversy — personally delivering a list of over 335,000 Napster usernames the band demanded be banned for trading its music. The case became a cultural flashpoint, casting Metallica as villains to many young fans even as it drew attention to a fundamental question: did artists have any control over how their work circulated online? The suit was ultimately settled out of court.

A&M Records, Inc. v. Napster, Inc. (2001). The more consequential case came from the major record labels themselves. A coalition led by A&M Records — and including what would become Sony, Warner, EMI, and Universal — sued Napster for contributory and vicarious copyright infringement. In February 2001, the Ninth Circuit Court of Appeals affirmed a preliminary injunction against Napster, finding that the platform had actual knowledge that infringing files were being shared and the practical ability to stop it, yet failed to do so. The ruling effectively forced Napster to shut down its original file-sharing service later that year, and the company filed for bankruptcy in 2002. Its brand was later revived as a licensed, paid streaming service — a fate not unlike the direction some now predict for AI companies: forced, eventually, into licensing deals with the rightsholders they once trained on without permission.

The parallel with Anthropic is instructive. Napster did not itself copy any songs — it built a system that let users do so. Anthropic, by contrast, is accused of doing the copying directly, at a massive scale, to build a commercial product. Courts historically have taken a dim view of both models when the underlying use was unauthorized, regardless of how transformative or technologically novel the platform claimed to be.

A Footnote on Reading the Fine Print

Every entertainment lawyer knows the story — perhaps apocryphal in its details, but true in substance — of Van Halen’s contract rider for their early-1980s stadium tours. Buried deep within pages of highly technical stage, electrical, and safety specifications was Article 126, requiring a bowl of M&M’s backstage with all the brown candies removed, on pain of forfeiting the full show fee. It wasn’t rock-star excess: it was a canary-in-the-coal-mine clause. If the band found brown M&M’s, they knew the promoter likely hadn’t read the technical specifications carefully either — and needed to re-inspect the entire rig for genuine safety hazards, such as staging that couldn’t bear the weight of their equipment. It is a fitting reminder, a generation later, that in copyright and entertainment law, the details buried in the fine print are rarely trivial — they are often where the real risk lives.

The Road Ahead

The Sony Music Publishing and Warner Chappell lawsuit will likely take years to resolve, but its trajectory looks increasingly familiar: courts distinguishing between the legitimate use of copyrighted material and the illegitimate means of acquiring it, mounting financial exposure for AI companies that trained on pirated content, and growing pressure toward licensing frameworks that compensate rightsholders. As with Napster, the technology itself is unlikely to disappear — but the terms on which it operates are being rewritten in real time, in courtrooms, not code.

This article is for general informational purposes and does not constitute legal advice. For questions about copyright and AI-related disputes, please contact Copyright Studio. hola@copyrightstudio.cat