Sony Music Publishing, Warner Chappell Music, and major music publishers filed a federal lawsuit in the U.S. District Court for the Northern District of California against Anthropic, CEO Dario Amodei, and co-founder Benjamin Mann, alleging a massive campaign of illegal torrenting, scraping, and downloading of copyrighted works to train the Claude AI model.
The Scale of the Allegations and the Legal Claims
- Direct infringement by torrenting against all three defendants.
- Contributory infringement by torrenting specifically targeting Amodei and Mann personally.
- Direct infringement through Claude’s lyric outputs.
- Digital Millennium Copyright Act (DMCA) violations for allegedly stripping copyright management information during scraping operations.
Statutory damages can reach up to $150,000 per willfully infringed work, alongside $25,000 per DMCA violation. When multiplied across tens of thousands of songs, the financial exposure climbs rapidly into the billions. Music copyright features a structurally layered architecture where a single track carries separate copyrights for its lyrics, musical composition, and sound recording—often held by entirely different entities, which multiplies potential claims.
Precedents and Previous Legal Battles
This litigation is not occurring in a vacuum. Anthropic has faced multiple intellectual property challenges regarding its model training datasets. In the Bartz v. Anthropic case, a group of authors accused the AI lab of utilizing copyrighted books to train its models. A judge ruled that while using copyrighted material for AI training could be legal, acquiring that content via piracy was not. Anthropic was subsequently ordered to pay $1.5 billion.
Furthermore, Concord Music Group and Universal Music Group filed a separate suit in January that has grown to cover more than 20,000 songs, seeking upwards of $3 billion. Other active plaintiffs in parallel music industry actions include ABKCO, BMG, and Round Hill Music.
The current Sony and Warner filing broadens the scope by accusing Anthropic of “flagrant piracy” through the use of shadow libraries like LibGen and Pirate Library Mirror to obtain millions of copies of books containing lyrics and sheet music.
Anthropic’s Defense and Industry Implications
Anthropic has responded to the litigation with a concise public statement. An Anthropic spokesperson stated, “We disagree with the publishers’ claims and we intend to defend ourselves robustly in court.”
This stance points toward a probable fair use or transformative use defense. Publishers counter that unlicensed AI training actively cannibalizes licensing markets that the music industry constructed.
The conflict places technical data scraping methodologies directly under judicial scrutiny. As large language model parameter scaling demands vast text and media corpora, the friction between AI labs and copyright holders continues to shape the legal boundaries of machine learning development.
