Sony and Warner Sue Anthropic: A Third Front Over Training Data

Sony Music Publishing and Warner Chappell Music, together with other music publishers, filed suit against Anthropic in late August. The claim: the company used tens of thousands of protected musical compositions without permission to train its Claude models. For the first time, all three major music groups are now suing the same AI company, and this case is far broader than the earlier ones.

Key takeaways

• Sony Music Publishing, Warner Chappell, and other publishers are suing Anthropic in the U.S. District Court for the Northern District of California.
• The complaint also names co-founders Dario Amodei and Benjamin Mann personally as defendants.
• It covers tens of thousands of song lyrics and compositions, which the plaintiffs say were obtained through torrenting, scraping, and bulk downloading.
• The publishers seek up to $150,000 per work plus up to $25,000 for each removal of copyright management information.
• Anthropic rejects the claims and says it will defend itself in court.

What the complaint alleges

The plaintiffs describe what they call a brazen campaign of illegally downloading, copying, and processing copyrighted works on a massive scale. Specifically, the complaint concerns lyrics and sheet music that Anthropic allegedly pulled from books and from freely available text collections online. A second allegation concerns copyright management information, meaning the data attached to a work identifying the rights holder, title, and license. The publishers say this information was stripped out when the data was ingested. Under U.S. law that is a separate cause of action, carrying up to $25,000 per instance.

The publishers request a jury trial and seek damages of up to $150,000 per infringed work. With tens of thousands of compositions involved, that adds up to billions on paper. Warner and Sony call the conduct one of the largest and most blatant ongoing thefts of intellectual property in history, a phrasing that mainly sets the tone of the dispute and decides little legally.

The core of the dispute is not legally new: it is the question of whether training an AI model on protected material counts as permissible use under U.S. case law or not. In earlier proceedings, Anthropic argued that the model learns from the texts without storing or reproducing them. The publishers counter that acquiring the data through unlicensed channels is the actual violation, regardless of what the model later outputs.

Why this case is bigger than the earlier ones

This is not Anthropic’s first time in court facing the music industry. Back in October 2023, Universal Music Publishing, Concord, and ABKCO sued over roughly 500 songs, mainly because Claude would reproduce protected lyrics verbatim on request. In January 2026, the same side filed a second suit that widened the scope to more than 20,000 works and demanded over $3 billion. The new Sony and Warner complaint raises the stakes again and, with the publishing arms of all three majors involved, now aligns the entire industry against Anthropic.

What is also new is that the plaintiffs name company leadership personally. Having co-founders and executives appear as named defendants is a signal: the publishers want not just money but also a look at how decisions were made inside the company. The recent dispute between German collecting society GEMA and music generator Suno likewise showed that courts now scrutinize how training data is acquired, as the Munich ruling against Suno demonstrated.

Context: Anthropic and its training data

For Anthropic this is the next front in a run of copyright disputes. Last year the company settled a class action by book authors for $1.5 billion, the largest known settlement of its kind so far. That case turned on Anthropic having sourced books from so-called shadow libraries. The lesson from it, that where the data comes from matters in the end and not just how it is used, shapes the current fight as well. A separate copyright suit against a rival, the case brought against OpenAI, raises the same underlying question.

Anthropic pushes back. In an initial statement the company said it disagrees with the publishers’ claims and intends to defend itself robustly in court. On the substance, meaning the question of which data sources were actually used, the company has so far said nothing.

Outlook

Whether the case reaches a jury or ends in a settlement like the authors’ suit is open. What is clear is that the music industry has sharpened its approach: more works, higher demands, personal defendants, and a second cause of action over stripped copyright information. For the AI industry as a whole, the math shifts. Training data was long treated as a cheap raw material that could simply be skimmed off the web. Every further lawsuit on this scale makes clear that the assumption can get expensive, and that licensing deals with rights holders are no longer a side issue for AI providers but part of the base cost.