Sony, Warner sue Anthropic over song use

Sony Music Publishing and Warner Chappell Music have filed a federal lawsuit against Anthropic in the U.S. District Court for the Northern District of California. The complaint targets the company behind the Claude AI models and names co‑founders Dario Amodei and Benjamin Mann as defendants. The publishers are pursuing a jury trial and seek damages and other relief tied to the alleged use of their catalogs in AI training.

The plaintiffs state that Anthropic harvested “thousands upon thousands” of copyrighted musical compositions without authorization. They allege that these works were incorporated into training data for Claude, citing specific, well‑known songs as examples of material they say appear in the datasets.

The complaint describes multiple channels through which Anthropic allegedly obtained protected material. It claims that lyrics and other works were acquired via torrenting, large‑scale scraping of websites including licensed lyric services, scanning of physical books, and reliance on large public datasets such as Common Crawl, The Pile, and Books3. These methods are presented as part of a broader pattern of large‑volume content collection for AI training.

The filing further alleges that co‑founder Benjamin Mann used BitTorrent to download millions of pirated books. It also claims that Anthropic employees accessed additional content from pirate archives, including Pirate Library Mirror. These details are cited as examples of how the company is alleged to have sourced copyrighted material beyond music.

A significant component of the lawsuit concerns the handling of copyright‑management information. The publishers allege that Anthropic removed identifiable copyright information from works during preprocessing of training data. They contend this occurred alongside the large‑scale acquisition of lyrics and other content used to train Claude models.

In terms of remedies, the complaint seeks statutory damages of up to $150,000 for each allegedly infringed musical composition. It also requests up to $25,000 for each instance in which copyright‑management information was allegedly altered or removed. The publishers argue that the potential damages are substantial given the volume of works they say were involved.

Beyond monetary relief, the lawsuit asks the court to compel disclosure of Anthropic’s training sources and methods. The publishers seek information on how training data was collected, processed, and integrated into Claude models, as well as any systems used to manage or track copyrighted material. These requests are positioned alongside the demand for a jury trial to resolve the claims.

Taken together, the complaint challenges both Anthropic’s alleged data acquisition practices and its internal handling of rights information. The case places the company’s approach to building large‑scale AI models under legal scrutiny and raises questions about how musical works and other copyrighted content have been used in the development of Claude.