Sony Music Publishing and Warner Chappell have filed a lawsuit against AI developer Anthropic in Northern California, accusing the company of incorporating copyrighted song lyrics into its Claude language models without authorization. The publishers demand up to $150,000 per infringed composition, highlighting ongoing tensions over copyright and AI training practices.
- Sony Music and Warner Chappell sue Anthropic for copyright infringement
- Litigation includes individual executives and claims up to $150,000 per song
- Munich court precedent rules memorization of lyrics by AI models is unlawful reproduction
What happened
Sony Music Publishing and Warner Chappell initiated legal proceedings in a Northern California court against Anthropic, alleging the unauthorized use of copyrighted song lyrics in training its Claude AI models. The lawsuit targets specific major hits by artists such as Mariah Carey, Taylor Swift, and others, claiming the lyrics were sourced from pirate archives like Library Genesis and Pirate Library Mirror.
Named personally in the suit are Anthropic executives Dario Amodei and Benjamin Mann. The plaintiffs are seeking statutory damages of up to $150,000 per infringing composition, an amount reflecting the maximum penalty allowed for willful copyright infringement under U.S. law, contrasting with lower settlements from past cases involving authors.
Why it matters
This lawsuit marks a critical escalation in how music publishers are confronting AI companies over training data sourcing. It follows a precedent set by a Munich court ruling in late 2025 that found the replicating of song lyrics within AI models constitutes reproduction, not covered by text and data mining exceptions, especially when sourced from unauthorized repositories.
The case highlights a widening gap in copyright enforcement approaches between the U.S. and Europe. While European regulations now require AI providers to document and disclose training data transparently under the AI Act, U.S. rights holders have to resort to litigation to uncover infringement, underscoring global legal and regulatory asymmetries.
What to watch next
The outcome of this case will be closely monitored as it could set a benchmark for liability concerning AI training data and copyrighted content. With damages sought reaching statutory maxima per song, the financial and operational implications for AI model developers could be substantial, influencing future data practices in the industry.
Additionally, the evolving regulatory landscape, notably in Europe with its AI Act and enforcement mechanisms, may pressure U.S. policymakers to consider more robust copyright transparency rules for AI. Stakeholders should watch for further legal interpretations related to AI data mining exceptions and potential new guidelines from enforcement bodies in both jurisdictions.