The essentials
Claude reproduces copyrighted lyrics verbatim; Anthropic defends AI training as fair use while plaintiffs seek $150,000 per work.
Technical evidence shows Claude AI models can generate copyrighted song lyrics verbatim, according to claims in ongoing litigation. Plaintiffs are seeking up to $150,000 in damages per infringed work in cases filed in San Jose, California.
Anthropichas mounted a fair use defense, arguing that training AI systems on copyrighted material falls within permissible use under copyright law. The defense challenges the plaintiffs' interpretation of how copyright protections apply to large language model development and deployment.
Other ways to read this story.
Alternative framings generated by SigNews. These are not independent sources or attributed quotations.
Verbatim lyric reproduction demonstrates specific copyright violation
The documented ability of Claude models to generate copyrighted lyrics verbatim provides concrete evidence of copyright infringement claims. This technical capability shifts the dispute from abstract concerns about data sourcing to measurable output that directly reproduces protected works without modification.
Fair use defense contests whether AI training qualifies as permitted use
Anthropic's invocation of fair use defense fundamentally disputes whether copyright restrictions apply to training data acquisition and model development. The defense argues existing copyright law accommodates this use, while plaintiffs' $150,000 per-work damages claim presumes fair use does not apply to these practices.