The battle over what artificial intelligence companies can use to build their models just hit another major industry. Sony Music Publishing and Warner Chappell Music have sued Anthropic, accusing the company behind Claude AI of unlawfully obtaining copyrighted songs and using the material to train its technology.
The Anthropic music lawsuit was filed in federal court in Northern California and also names Anthropic CEO and co-founder Dario Amodei and co-founder Benjamin Mann as defendants. The publishers allege Anthropic obtained copyrighted lyrics and sheet music through methods that included torrenting, scraping websites, and downloading material from unauthorized sources before using copyrighted works in connection with Claude’s development.
That distinction could become critical.
The fight over AI training has increasingly centered on two separate questions: whether training an AI model on copyrighted material can qualify as fair use, and whether the way a company obtained that material was lawful in the first place. Sony and Warner’s publishers are putting heavy emphasis on the second issue.
According to the complaint, the publishers accuse Anthropic of a “brazen campaign of illegally torrenting, scraping, and downloading copyrighted works.”
The catalogs potentially involved are not exactly small-time material. Reporting on the lawsuit identifies works associated with The Beatles, Taylor Swift, Michael Jackson, Mariah Carey, Marvin Gaye, Bruno Mars, and other major artists among material cited by the publishers.
Songs referenced in reporting on the complaint include “All I Want for Christmas Is You,” “Ain’t No Mountain High Enough,” “Eye of the Tiger,” “Hallelujah,” “Uptown Funk,” “I Am the Walrus,” “Paper Rings” and “Cruel Summer.”
The publishers also allege Claude can reproduce copyrighted lyrics in its outputs. They argue that Anthropic not only copied protected works during the development process but created technology capable of generating material that could compete with the copyrighted works themselves.
Anthropic disputes the accusations.
“We disagree with the publishers’ claims, and we intend to defend ourselves robustly in court,” an Anthropic spokesperson said.
The potential price tag makes this more than another copyright disagreement floating through Silicon Valley.
The publishers are seeking statutory damages that could reach $150,000 for each work found to have been willfully infringed. Reporting on the complaint says they are also seeking damages tied to alleged removal of copyright-management information. With potentially thousands or tens of thousands of compositions at issue, the financial exposure could climb into the billions depending on what claims ultimately survive and what a court finds.
No court has ruled that Anthropic committed the alleged infringement in this new case.
Still, the lawsuit arrives with some serious legal history already sitting in the background.
Anthropic previously faced litigation from authors who accused the company of using pirated books in connection with AI training. In that litigation, U.S. District Judge William Alsup drew an important line between AI training itself and how training material was acquired.
Alsup found that Anthropic’s use of books for training was transformative fair use under the circumstances before him. However, the company’s acquisition and retention of millions of pirated books created a separate copyright problem. Anthropic later reached a $1.5 billion settlement with authors over piracy-related claims.
Now the music publishers are trying to keep the spotlight on acquisition.
The latest case also isn’t Anthropic’s first clash with the music business. Universal Music Publishing Group, Concord Music Group and ABKCO sued Anthropic in 2023 over allegations involving copyrighted song lyrics. Other music publishers have since pursued additional claims as the industry’s legal confrontation with AI companies has expanded.
For music companies, the issue reaches beyond whether an AI chatbot occasionally spits out familiar lyrics. Publishers make money by licensing compositions, lyrics, and other copyrighted material. Their lawsuits argue that AI companies shouldn’t get to bypass those licensing systems simply because copyrighted works are being fed into a massive technological pipeline.
AI companies, meanwhile, have argued in copyright litigation that training models can constitute transformative fair use.
That legal question is becoming one of the biggest fights surrounding generative AI, and the stakes extend far beyond Claude. Authors, news organizations, artists, music companies and other copyright owners have brought cases challenging how technology companies source and use creative work to train artificial intelligence.
The timing adds another layer. On September 2, the U.S. government urged G20 nations to adopt frameworks allowing AI companies to train models on copyrighted content while protecting creators’ rights. The Trump administration has also backed OpenAI’s fair-use position in its separate copyright battle with The New York Times.
That means courts, policymakers, creators, and some of the world’s richest technology companies are all wrestling with the same question at once: How much copyrighted human work can go into building AI before somebody has to pay?
Sony and Warner are making their position clear.
Now Anthropic has to make its case in court.
