- Lead. Sony Music Publishing and Warner Chappell Music filed a 48-page federal copyright lawsuit against Anthropic, CEO Dario Amodei and co-founder Benjamin Mann on August 29, 2026, alleging the company conducted a “brazen campaign of illegally torrenting, scraping and downloading” copyrighted song lyrics and sheet music to train its Claude large language models.
- Fact. The complaint, lodged in the US District Court for the Northern District of California, identifies specific works allegedly used without permission, including Mariah Carey’s “All I Want for Christmas Is You,” Survivor’s “Eye of the Tiger,” songs by Taylor Swift, The Beatles and Bon Jovi, and seeks statutory damages of up to $150,000 per infringed copyright.
- Stake. With potentially tens of thousands of works at issue, total liability could reach the billions — a figure consistent with an earlier $1.5 billion settlement Anthropic reached in the Bartz v. Anthropic class action, where a judge found that while training on copyrighted content may be defensible, acquiring it through piracy is not.
The lawsuit follows earlier actions filed by the same law firm on behalf of Universal Music Publishing Group and Concord Music Group in January 2026, which covered more than 20,000 works and sought more than $3 billion in damages. The Sony and Warner complaint adds two of the world’s most commercially significant music publishers to the list of plaintiffs pursuing Anthropic over alleged training data theft.
What the Publishers Allege
According to the filing, Anthropic’s training pipeline drew on unauthorised sources including torrent sites, the Common Crawl archive and the pirate database Library Genesis — all of which the publishers describe as known repositories of infringing content. The complaint characterises Anthropic’s conduct as systematic rather than incidental, noting that the company reportedly prioritised acquiring large-scale text corpora at speed without establishing licensing arrangements that competitors in the music industry have pursued.
Sony Music Publishing and Warner Chappell represent the publishing rights — covering lyrics and compositions — rather than the master recordings, which are the domain of recorded-music divisions such as Sony Music Entertainment and Warner Music Group. Publishing rights carry their own copyright protections and their own statutory damage schedules under US law.
Anthropic’s Position
In a statement quoted across multiple outlets, an Anthropic spokesperson said: “We disagree with the publishers’ claims and we intend to defend ourselves robustly in court.” The company has not filed a detailed response in the litigation. Anthropic has argued in other contexts that training large language models on publicly accessible text is a transformative use that falls within fair-use doctrine — a legal theory that has not yet been definitively tested at the appellate level in the United States.
A Pattern Across the Industry
The Anthropic case is part of a broader wave of AI copyright litigation involving training data. OpenAI faces similar suits from book authors and news organisations; Google and Meta have also been named in related actions. What distinguishes the Anthropic music cases is the specificity of the piracy allegation: the publishers are not arguing simply that training on copyrighted lyrics is infringement, but that Anthropic acquired the material through channels it knew to be infringing.
That distinction proved decisive in the Bartz settlement, and it is likely to be the central legal question in the Sony and Warner action as well. Anthropic raised $3.5 billion from Google and other investors earlier this year; the legal reserve implications of a broad adverse ruling would be significant for a company still burning cash at scale in its race against OpenAI and Google DeepMind.