Anthropic Sued by Sony and Warner Music Over Claude's Training Data

Dario Amodei built Anthropic on the promise that it was the "safe" AI lab, the one that would do things differently. Now he's named personally in a lawsuit that could cost billions of dollars — and it's not about a chatbot going rogue, it's about song lyrics.

What Sony and Warner are actually accusing Anthropic of

Sony Music Publishing and Warner Chappell Music filed suit against Anthropic in federal court in Northern California, and they didn't just name the company. Co-founders Dario Amodei and Benjamin Mann are listed as individual defendants too, a move that signals the publishers want to hold the people running Anthropic personally liable, not just the corporate entity that can absorb a settlement and move on.

The complaint says Anthropic illegally torrented, scraped, and downloaded copyrighted songs by the tens of thousands to train its Claude models, without ever licensing the material. The list of works named includes "Eye of the Tiger," Marvin Gaye's "Ain't No Mountain High Enough," Taylor Swift's "Paper Rings," and even Christmas standards like "All I Want for Christmas Is You." The publishers are seeking statutory damages (a fixed dollar amount per violation that courts can award without the plaintiff having to prove exact financial loss) of up to $150,000 per song. With tens of thousands of compositions in play, the exposure could run into the billions.

"One of the largest and most blatant ongoing thefts of intellectual property in history."

That's the language from the complaint itself, and it's deliberately harsh — publishers know these filings get quoted in headlines, and they're framing this as a moral case as much as a legal one.

This isn't Anthropic's first copyright fight

Anthropic already settled a $1.5 billion lawsuit with book authors over pirated books used in training data, after a judge ruled that training an AI model on copyrighted books could count as fair use, but that acquiring those books through piracy sites was a separate, illegal act. That distinction is exactly what the music publishers are leaning on this time: it's not really an argument about whether AI can learn from copyrighted material, it's an argument about how the material was obtained in the first place.

A few things this case will likely turn on:

• Whether courts keep separating "training is fair use" from "how you got the training data"
• Whether naming founders personally becomes a standard tactic in future AI copyright suits
• Whether a damages number this large forces AI labs to license music catalogs upfront rather than fight it out later

Why this matters beyond Silicon Valley — the India angle

India has its own version of this exact fight already playing out. News agency ANI sued OpenAI in the Delhi High Court over ChatGPT allegedly being trained on its copyrighted articles without permission or payment, and the case has dragged on precisely because Indian copyright law has no clear answer for AI training. The Copyright Act, 1957 doesn't include anything like a text-and-data-mining exception (a specific carve-out some countries use to let AI systems process copyrighted content for training purposes), so Indian courts are being asked to figure this out from first principles, the same way American courts are now doing with Anthropic.

For India's own music industry — T-Series, Saregama, Zee Music and the rest sit on catalogs worth billions of dollars — this US lawsuit is a preview. If Sony and Warner win big, expect Indian labels to start asking the same questions about how foreign AI companies have used Bollywood soundtracks and regional-language music to train models sold right back into the Indian market. And for Indian AI startups building their own models, it's a warning that scraping first and licensing later is getting more expensive by the month.

What happens next

Anthropic has pushed back, with a spokesperson saying the company disagrees with the publishers' claims and intends to "defend ourselves robustly in court," pointing back to the same fair-use reasoning a judge accepted in the book-authors case. But Sony and Warner aren't the first music rights holders at Anthropic's door — Universal Music Publishing, Concord, and ABKCO sued back in 2023, the same group filed again in January 2026 over more than 20,000 additional works, and BMG brought its own case in March. This is now a pattern, not an isolated dispute, and a case this size will take years to resolve through discovery and likely appeals. The direction of travel, though, is clear: AI companies that once treated the internet as free training material are now negotiating licensing deals with publishers, or getting sued by the ones they didn't. The music industry watched what happened with Napster two decades ago, and it isn't interested in repeating that mistake with AI — this time, it's suing before the damage is a fait accompli, not after.