Music Publishers Sue Anthropic Over Claude Training Data

The complaint landed in a federal courthouse late on a Thursday evening, and it named not just a company but two of its founders. Sony Music Publishing and Warner Chappell, joined by a group of music publishers, accuse Anthropic of building its Claude models on millions of pirated books, including song lyrics and sheet music, gathered through illegal torrenting and web scraping. Filed Aug. 28 in the Northern District of California, the lawsuit describes the conduct as “an unbridled campaign of piracy.” Music Business Worldwide first reported the case.

The defendants include Anthropic’s chief executive, Dario Amodei, and co-founder Benjamin Mann. The publishers say the two executives directed a data-collection operation that treated copyright as an obstacle to be routed around, and they are seeking damages plus an injunction against further use of the material. The complaint argues that the scale of the alleged taking sets it apart from the disputes that preceded it: millions of books, scraped from the open web and shared through peer-to-peer networks, then folded into the training runs that produced Claude.

Anthropic pushed back in a statement. “We disagree with the allegations and will defend ourselves vigorously in court,” a spokesman said.

The case arrives at a delicate moment. Anthropic is in the middle of an initial public offering roadshow built around a valuation that people familiar with the plans have put at $2 trillion. Amodei spent part of the weekend on Wall Street, appearing alongside Salesforce’s chief executive, Marc Benioff, according to people familiar with the matter. A copyright suit filed in the middle of that process hands underwriters and potential investors another question to price, and it gives the company’s critics a concrete story to tell about how its models were built.

It is also not the first. Concord Music Group and Universal Music Group sued Anthropic in January over similar claims. And a federal judge this year awarded authors $1.5 billion in the Bartz class action, a case brought by writers who said their books had been used to train the models without permission. The reasoning in the Bartz case cut both ways: training an AI system on copyrighted works is itself lawful, the judge held, but obtaining those works through piracy is not. That distinction is now the battleground. The publishers in the new suit argue Anthropic crossed the line on the second half of that equation, using torrent sites and scrapers rather than licensed sources.

The music industry’s posture toward AI is more complicated than simple opposition. Universal, Sony and Warner all took stakes in Stability AI this year, betting that generative models will eventually need licensed creative material. The same companies now suing Anthropic over its training data are investors in the broader AI economy. The dual track, investing in the winners while suing the outliers, lets the industry protect itself either way the market moves.

Lawyers who follow the cases say the outcome could set the price of training data across the industry. If the publishers win, licensing negotiations for lyrics, sheet music and books shift in favor of rightsholders, and every AI lab faces a higher bill for the material its models learn from. If Anthropic prevails, the open-web approach to data collection gets a judicial green light that other labs would be free to copy. Either result would echo far beyond music, because the same legal questions govern text, images and code.

For Anthropic, the stakes go beyond a single suit. The company has positioned itself as the safety-conscious alternative in AI, publicly committed to responsible development and to compensating creators through programs that pay publishers whose work appears in training data. A trial record that includes an accusation of systematic piracy complicates that story, whatever the verdict.

The publishers asked the court for damages and a permanent injunction. They also requested a jury trial, which in a copyright case can mean the dispute runs for years. Anthropic said it will respond once it has reviewed the complaint.

More suits may be coming. Copyright holders in music and publishing have watched the Bartz award and the January Concord case and drawn the same conclusion: courts are willing to hear these claims, and the damages can be enormous. The complaint names songs, scores and books by named artists and writers, giving the case a concrete catalog that will be hard to dismiss on procedural grounds.

The industry is watching the roadshow, too. Anthropic has said publicly that AI models should respect creators and that it supports paying for the material that trains them. The publishers’ lawyers say the company’s actions did not match those words. A jury will eventually sort out which account is true, but the argument itself will shape how the next round of deals between AI labs and the music business gets negotiated.

Related Posts

  • September 6, 2026
  • 7 views
Anthropic Moves Its IPO Filing to Late September

The bankers and lawyers running Anthropic’s initial public offering had told investors to expect the company’s registration documents as soon as this week. The calendar has moved. Anthropic now plans…

  • September 6, 2026
  • 6 views
OpenAI Quietly Revises GPT-6 Astra Scores After Launch

When OpenAI released GPT-6 Astra on Sept. 3, the launch post carried the usual furniture of a modern model debut: coding results, speed comparisons and a figure for how often…