Two of the world's largest music publishers — Sony Music Publishing and Warner Chappell — filed a joint copyright suit against Anthropic on Friday, naming co-founders Dario Amodei and Benjamin Mann individually alongside the company. The suit alleges that Anthropic engaged in a "brazen campaign of illegally torrenting, scraping, and downloading" copyrighted works — specifically lyrics and sheet music embedded in books used to train Claude — and frames that acquisition as independently actionable, regardless of whether using copyrighted material in AI training ultimately constitutes infringement.

The legal theory advanced in the filing is meaningfully different from the copyright suits that have proliferated since the generative AI era began. Most of those suits argue over whether AI training constitutes infringement at the point of use — a question that courts in the United States have not yet resolved consistently, and on which the balance of judicial opinion remains unsettled. This suit argues that Anthropic's liability is established at the point of acquisition: that the company pirated works it did not have the right to copy, and that the piracy itself is actionable under copyright law, independent of what the works were subsequently used for.

That distinction matters because a January 2026 precedent ruling established that piracy-based acquisition of training data is not protected by fair use defences, even when those defences might apply to the training use itself. If that ruling is applied to this case, Anthropic cannot rely on the fair-use-for-training arguments it and other AI companies have developed; it must instead contest whether the acquisition methods employed during its training data assembly constituted piracy. That is a factually narrower question — one that hinges on what Anthropic's data engineers actually did, not on interpretations of copyright doctrine — and therefore potentially easier for plaintiffs to win.

Anthropic has stated that it "disagrees with the publishers' claims" and intends to "defend ourselves robustly in court," without addressing the piracy theory specifically. The damages amount has not been disclosed in public filings.

The case arrives at a moment when Anthropic is simultaneously defending a separately filed suit from book publishers, managing the Pentagon supply-chain-risk ruling in its favour, and navigating the operational demands of a company that has grown substantially in revenue and headcount over the past eighteen months. It is also the second large copyright settlement the company faces: a 2024 suit — the Bartz case — ended in a court-ordered settlement of one point five billion dollars after a ruling that Anthropic had used copyrighted content in training without sufficient licensing.

The music industry's decision to pursue the piracy theory, rather than the simpler training-use theory that the Bartz case initially deployed, suggests that publishers are adapting their litigation strategy based on what has worked and what has not. The Bartz settlement produced a significant payment but did not resolve the underlying training-use question as a matter of law. The piracy theory, if it succeeds, would establish that frontier labs have a pre-training obligation to document and justify how every work in their training datasets was acquired — a compliance requirement that would affect every large-scale training operation in the industry, not only Anthropic's.

The music industry has moved faster than other creative industries throughout the AI copyright disputes. Book publishers, news organisations, and visual artists have all pursued parallel litigation, but the music publishers have been the most aggressive in developing legal theories that reach further up the value chain. A ruling against Anthropic on piracy-based acquisition would be the most significant legal development in AI copyright law since the January 2026 precedent, and would create pressure for settlements — or dramatically more careful data provenance documentation — across the entire frontier model development ecosystem.