What happened
Sony Music Publishing, Warner Chappell and numerous other music publishers sued Anthropic, plus co-founders Dario Amodei and Benjamin Mann. The complaint was filed on Friday evening in federal court in Northern California. Music Business Worldwide reported it first.
The publishers' accusation is blunt. They say Anthropic ran a "brazen campaign" of torrenting, scraping and downloading copyrighted works without permission, and that thousands of those works went into training Claude. They also use the phrase "blatant theft", which is not the language of people hoping for a quiet settlement call.
Anthropic's answer was short. A spokesperson told TechCrunch the company disagrees with the claims and intends to defend itself "robustly" in court.
This is not Anthropic's first trip to this particular rodeo. Several of the same lawyers already act for Concord and Universal Music Group in a separate suit from January. They also led Bartz v. Anthropic, where authors accused the company of using copyrighted works to train products like Claude.
Why it matters
Naming the co-founders personally is the spicy part. Suing a company is routine. Putting two executives' names on the caption says the plaintiffs want the people at the top to feel it.
The core allegation is also about how the data was obtained, not only what the model learned. Torrenting is a different argument from "we read the open web". Courts tend to care about that difference, and so will every other lab with a data pipeline it would rather not describe in detail.
Add the January case and the authors' fight, and a pattern shows up. The same legal talent keeps finding new rights holders to represent against the same defendant. Training data is turning from a footnote into a line item.
What it means for Europe
The suit is American, but the question travels. Training on protected works without clearance is a live legal risk for AI developers everywhere, including under EU intellectual property rules. European publishers and labels will read this complaint closely, and European AI teams should assume their own data provenance gets the same scrutiny.
What to do with it
- Audit your sources.If you build or fine-tune models, write down where every training dataset came from before someone asks you in court.
- Read your vendor terms.Check what your AI provider says about copyright indemnity and what happens if a training-data claim hits them.
- Watch the docket.Follow how Anthropic's defense and the earlier cases develop, because they will set the tone for everyone's data policy.
- Creators, document your work.If you publish music or writing, keep dated records of ownership and check how your catalog is licensed.
Ciao, sapiens. The models read everything, and now the lawyers are reading back.


