Sony, Warner Music Sue Anthropic for Copyright Infringement Over Songs Used in Claude AI Training
The music industry’s uneasy standoff with generative AI companies just escalated significantly. Sony Music Publishing and Warner Chappell Music filed a joint federal lawsuit against Anthropic on Friday, accusing the AI company of illegally scraping, torrenting and downloading tens of thousands of copyrighted songs to train its Claude chatbot models. The complaint, lodged in the U.S. District Court for the Northern District of California, names Anthropic itself along with CEO Dario Amodei and co-fo
under Benjamin Mann as defendants, and it doesn’t hold back on language, describing the alleged conduct as one of the largest and most blatant ongoing thefts of intellectual property in history.
According to the 48-page filing, Anthropic pulled lyrics and sheet music from artists ranging from The Beatles and Michael Jackson to Taylor Swift and Mariah Carey, drawing on pirate sources including torrent networks, Library Genesis, and datasets like Common Crawl rather than licensing the material properly. The publishers allege the company went further than simply feeding this content into training data. They claim Claude was also capable of reproducing verbatim lyrics in response to user prompts, effectively competing with the original songs it had allegedly lifted from, and that Anthropic stripped identifying metadata from tracks during processing, cutting songwriters out of proper credit and attribution in the process.
Sony and Warner are asking for a jury trial and seeking the maximum statutory damages available under copyright law, up to $150,000 for every infringed composition, plus an additional $25,000 for each instance where copyright management information was allegedly altered or removed. Given that the suit describes thousands upon thousands of copyrighted works being harvested, the potential damages could run into the billions if the publishers prevail.
What makes the timing notable is that this isn’t Anthropic’s first brush with music industry litigation, or even its first copyright fight overall. Universal Music Group sued the company back in 2023 and again earlier this year over similar allegations involving song lyrics, and both of those cases remain active. Concord Music Group and Universal jointly filed another suit in January seeking more than $3 billion in damages tied to claims that Anthropic illegally downloaded more than 20,000 copyrighted songs. BMG has its own pending case as well, though a narrower one, covering 493 compositions compared to the far broader scope Sony and Warner are alleging here.
Anthropic’s more consequential legal exposure, though, traces back to a separate case brought by a group of authors, known as Bartz v. Anthropic, which resulted in the company agreeing to pay $1.5 billion last year to settle claims that it trained its models on pirated books. The judge overseeing that case delivered a ruling that’s become central to how these disputes are now being argued on both sides: training AI models on copyrighted material can qualify as transformative fair use, but obtaining that material through piracy in the first place does not get the same protection. Sony and Warner’s complaint leans directly on that distinction, arguing that regardless of how fair use might apply to the training itself, the manner in which Anthropic allegedly acquired the songs was unlawful from the start.
The publishers didn’t shy away from invoking that earlier settlement in their filing, arguing that a payout of that size clearly wasn’t enough to change Anthropic’s behavior. Sony and Warner wrote that the company “clearly considers that to be just the cost of doing business,” pointing to Anthropic’s roughly $2 trillion valuation as evidence that the previous penalty did little to discourage what they describe as a business model still built on copyright theft.
Anthropic, for its part, isn’t backing down either. A company spokesperson pushed back on the lawsuit, characterizing it as the third case brought by the same group of lawyers and arguing it largely recycles allegations already sitting before the courts in other active disputes. The spokesperson reiterated Anthropic’s long-standing position that training generative AI models qualifies as transformative fair use, referencing the Bartz ruling directly, and said the company intends to defend itself robustly against the new claims.
This case lands at a moment when the broader relationship between the music industry and AI companies has been anything but settled, oscillating between aggressive litigation and quiet licensing negotiations depending on which side currently has more leverage. Labels and publishers have generally pursued lawsuits first, using the threat of massive statutory damages to pull AI companies to the negotiating table, before eventually striking licensing deals once liability becomes clearer or too costly to keep fighting. Whether Sony and Warner are angling for a similar outcome here, or genuinely intend to take Anthropic through a full trial, remains to be seen, but the sheer scope of this complaint, described by legal observers as notably broader than the narrower BMG case, suggests the publishers see leverage in going big rather than settling early.
For Anthropic, the timing adds another layer of complexity to a period already shaped by regulatory scrutiny and rapid commercial growth. The company has positioned itself as a leader in enterprise and developer-facing AI tools through its Claude models, and a prolonged legal battle over training data provenance could shape how it and its competitors approach content licensing going forward, regardless of how this particular case resolves. Readers can find Anthropic’s official statements on AI training and copyright at anthropic.com. Techchora will continue tracking how this case and the parallel Universal, Concord and BMG lawsuits progress as the courts work through what has become one of the defining legal questions of the generative AI era.
