Sony Music and Warner Chappell Sue Anthropic for Pirating Thousands of Copyrighted Songs for AI Model Training

Sony Music Publishing and Warner Chappell Music filed a lawsuit in a Northern California district court against AI firm Anthropic and cofounders Dario Amodei and Benjamin Mann, alleging the company illegally torrented and scraped copyrighted song lyrics to train its “Claude” AI models, seeking up to $150,000 per composition in statutory damages.

The Bottom Line

• The Training Pipeline: The lawsuit alleges Anthropic bypassed licensing fees by utilizing illegal digital archives like Library Genesis and Pirate Library Mirror to ingest thousands of copyrighted works.
• Market Precedent: This action builds upon a fragile legal landscape for large language model (LLM) developers, following Anthropic’s multi-billion-dollar settlement with authors and ongoing litigation from outlets like The New York Times Company (NYSE: NYT).

Unpacking the Complaint: Scale and Method

According to the court filing, Anthropic conducted what publishers termed a brazen campaign of scraping and downloading copyrighted works on a massive scale. The objective: operating and commercializing the “Claude” series of artificial intelligence models.

Among the thousands of tracks cited are 1980s anthems like “Eye of the Tiger,” Marvin Gaye’s “Ain’t No Mountain High Enough,” Mariah Carey’s holiday standard “All I Want for Christmas is You,” and Taylor Swift’s “Paper Rings.” Here is the math of the courtroom: statutory damages of $150,000 per composition across hundreds or thousands of works translate to an existential financial exposure for foundational model developers.

Data Provenance and the Shadow of Library Genesis

The music publishers stated that Anthropic harvested copyrighted compositions through digital archives including Library Genesis and Pirate Library Mirror. This builds directly upon a June 2025 judicial ruling establishing that Anthropic downloaded over 7 million pirated books to train Claude.

Those ingested volumes contained the lyrics and sheet music to hundreds of foundational musical compositions. Exhibit A of the lawsuit points to works such as “Livin’ On a Prayer,” “September,” “Great Balls of Fire,” “Ramblin’ Man,” and “Hallelujah.” But the balance sheet tells a different story regarding output: the publishers allege that Claude models generate identical or near-identical copies of these copyrighted works in response to user prompts, creating a direct substitute for human-created music.

Selected Copyright Infringement Allegations in Music Publishers v. Anthropic

Complainant / Entity	Target Company	Core Allegation	Financial Remedy Sought
Sony Music Publishing & Warner Chappell Music	Anthropic (Dario Amodei, Benjamin Mann)	Illegal torrenting and scraping of copyrighted lyrics via pirate archives to train “Claude” LLMs	Up to $150,000 per infringed musical composition in statutory damages

Broader Market Implications for Generative AI

This litigation arrives amid a broader wave of legal friction testing the boundaries of fair use in machine learning. Tech companies require immense troves of unstructured data to maintain model competitiveness, frequently pushing legal limits to secure training corpuses. Last September, Anthropic agreed to pay over $1.5 billion to authors to settle a class-action lawsuit concerning pirated books.

Simultaneously, competitors like OpenAI face parallel litigation from major media institutions, including The New York Times (NYSE: NYT) and Encyclopedia Britannica. As copyright holders test statutory damages in federal courts, the economic model of scraping unstructured web data faces mounting compliance costs. For enterprise adopters integrating LLMs, these legal headwinds threaten to force a structural repricing of AI inference and training models as licensing fees become mandatory line items.

Disclaimer: The information provided in this article is for educational and informational purposes only and does not constitute financial advice.