Sony Music Publishing and Warner Chappell Music have sued Anthropic in the U.S. District Court for the Northern District of California, alleging that the developer of Claude engaged in large-scale copyright infringement involving musical compositions controlled by the publishers. The lawsuit, reported as filed on August 28, 2026, seeks a jury trial and remedies that include statutory damages, destruction of allegedly infringing copies, and an accounting of Anthropic’s training data and methods.

Allegations in the complaint

The complaint’s allegations have not been proven in court. According to reporting on the filing, the publishers contend that Anthropic illegally torrented, scraped and downloaded copyrighted works as it developed and operated Claude. They characterize the alleged conduct as involving tens of thousands of their musical compositions, a scale that could make the damages request consequential if liability is ultimately established.

Sony Music Publishing and Warner Chappell Music are music publishers, meaning their claims concern the underlying musical works—the lyrics and compositions—rather than necessarily particular sound-recording performances. That distinction matters in music-rights disputes: a song can involve separate rights in the composition and in a recorded performance. The reported case is centered on the publishers’ composition catalogs and Anthropic’s alleged use of those works.

Reporting on the complaint says Anthropic co-founders Dario Amodei and Benjamin Mann are named as individual defendants alongside the company. The suit reportedly advances four claims, including direct and contributory infringement related to alleged torrenting, direct infringement, and removal or alteration of copyright-management information. Anthropic had not provided a public response in the reporting included in the source material; TechCrunch said it was unable to reach the company for comment before publication.

Damages and requested remedies

The monetary exposure sought is framed through statutory damages rather than a single stated total. The publishers reportedly request up to $150,000 per work for willful infringement, if willfulness and infringement are found, as well as up to $25,000 for each alleged removal or alteration of copyright-management information. Those are requested legal remedies, not an award, and any damages would depend on the claims that survive litigation, the evidence produced and eventual rulings by the court or a jury.

The request for destruction and for an accounting is also important beyond damages. If granted in some form, those remedies could force closer examination of what materials were obtained, how they moved through training and development systems, and whether allegedly infringing copies remain in datasets or related infrastructure. The case therefore raises questions not only about whether copyrighted works were used, but also about what a practical remedy looks like once material has been incorporated into a large-scale AI development process.

Questions around AI training

For AI developers, the case adds pressure to distinguish among several legally separate issues that are often collapsed into one public debate: how training material was acquired, whether copying occurred during dataset construction or training, whether an asserted use is protected by fair use, and whether a model can generate outputs that reproduce protected expression. Sony and Warner Chappell’s allegations span acquisition and use, rather than resting solely on isolated output examples. Anthropic will have an opportunity to respond in court, and the complaint itself is only the opening pleading in what could be a lengthy process.

The wider legal setting remains unsettled. In its May 2025 report on generative-AI training, the U.S. Copyright Office said that copyrighted works used in AI development should not be treated as ordinary “data,” because they embody protected creative expression. The office also described fair use as the principal defense where AI development activities amount to prima facie infringement, while emphasizing that the analysis depends on the particular facts and the overall use. That framework underscores why disputes involving source material, model behavior and market effects may turn on detailed evidence rather than broad claims about AI training in the abstract.

The Copyright Office’s report also found strong interest among rightsholders and creators in voluntary collective licensing for AI uses, while recommending that the licensing market be allowed to develop without government intervention for the time being. That policy context is especially relevant to music, an industry with established licensing systems but complex and fragmented rights ownership. The Anthropic litigation could increase commercial incentives for developers and rightsholders to negotiate permissions, data-access terms or other arrangements while courts continue to define the boundaries of infringement and fair use.

Related copyright litigation

A separate 2025 ruling in Thomson Reuters’ case against legal-AI company Ross Intelligence illustrates that courts can closely examine the nature of copied material and the function it serves in a product. That decision did not involve generative music or Claude, so it does not resolve the publishers’ claims against Anthropic. But it reinforces a broader point: arguments about technical innovation do not eliminate the need for courts to evaluate the specific copyrighted expression, the purpose of the copying and the alleged market impact.

What comes next

For now, the Sony Music Publishing and Warner Chappell case is an allegation-driven legal challenge, not a judicial finding about Claude or Anthropic’s practices. Its significance lies in the scale alleged by two major music publishers and in the remedies sought. As the case proceeds, the central questions will be whether the publishers can substantiate their account of acquisition and use, what defenses Anthropic raises, and whether the court’s treatment of training-related copies and generated material offers guidance for other generative-AI copyright disputes.