Sony Music Publishing, Warner Chappell, and several music publishers have filed a lawsuit against Anthropic and its founders, Dario Amodei and Benjamin Mann, accusing the artificial intelligence lab of carrying out a wide-ranging campaign to unlawfully obtain copyrighted works and then use them to train the Claude model.
The lawsuit was filed late Friday in the U.S. District Court for the Northern District of California. According to the plaintiffs, Anthropic’s practices included torrenting, scraping, and unauthorized downloading of thousands of copyrighted works. TechCrunch was unable to obtain comment from the company before publication.
Accusation Goes Beyond Mere Use of Content in Training
The case is not limited to the usual question of whether training artificial intelligence models on protected material constitutes lawful use. The lawsuit accuses Anthropic of unlawfully obtaining millions of copies of books through torrenting, including books containing song lyrics and musical notation.
The publishers describe what happened as “blatant theft” and “egregious piracy.” However, these descriptions represent the plaintiffs’ allegations in the lawsuit and do not constitute a final judicial ruling against Anthropic. The article also contains no indication of the company’s legal position or response to the accusations.
Connection to Earlier Disputes
The new lawsuit comes amid a series of cases Anthropic is facing over intellectual property. Some of the attorneys involved also represent Concord Music Group and Universal Music Group in a case filed in January, and they have been associated with Bartz v. Anthropic, in which authors accused the company of using copyrighted books to train products such as Claude.
According to the article, Anthropic was ordered to pay $1.5 billion in the Bartz case after a judge concluded that using copyrighted works to train artificial intelligence systems was lawful, but obtaining those works through piracy was not lawful.
Why Does This News Matter?
The current lawsuit adds a more specific layer to the dispute over artificial intelligence training data: even if courts accept the principle that using copyrighted content in training is permissible under certain circumstances, the source of the data and the method by which it was obtained may create separate legal liability. This puts model publishers before a practical question concerning training-source records and procedures for obtaining content, rather than merely the model’s design or outputs.
The outcome in this case remains open; the available material contains no response from Anthropic and no new court ruling. The scope of the lawsuit, which includes books containing song lyrics and musical notation, also makes it different from earlier cases in terms of the types of material rights holders seek to protect.