Anthropic, the maker of the Claude AI chatbot, has reached an agreement with major music publishers to settle parts of an ongoing copyright infringement lawsuit. The deal, signed off by U.S. District Judge Eumi Lee, outlines measures to address concerns over the unauthorized use of copyrighted song lyrics.
Key Points:
- Anthropic agrees to maintain existing guardrails preventing its Claude AI from reproducing or creating derivative works from publishers' copyrighted lyrics
- The agreement establishes a formal process for publishers to flag potential copyright violations
- This case represents the first legal action by music publishers against an AI company over lyrics used in large language model training
The lawsuit, filed in Tennessee federal court, marks a significant legal battle over AI and intellectual property rights. Music publishers accused Anthropic of using lyrics from over 500 songs in its AI training data and alleged that Anthropic's Claude chatbot provided near-identical reproductions of copyrighted lyrics from songs like Katy Perry's "Roar" and works by The Rolling Stones and Beyoncé.
At the heart of the case is the argument that AI systems like Claude undercut established markets for licensed lyrics, such as lyric aggregation websites. While the agreement doesn’t end the lawsuit, it establishes a protocol for publishers to intervene when they suspect copyright breaches.
Anthropic emphasized that its chatbot isn’t designed for copyright infringement and highlighted its existing technical guardrails aimed at preventing the reproduction of copyrighted content. The company reiterated its position that using copyrighted material in AI training constitutes 'fair use' under existing laws.
The settlement's framework allows music publishers to notify Anthropic if they believe the guardrails aren't effectively preventing copyright infringement. Upon receiving such notification, Anthropic is required to conduct a prompt investigation, with both parties agreeing to cooperate in good faith.
While this agreement resolves some immediate concerns, the court is still expected to rule in the coming months on a broader preliminary injunction that would restrict Anthropic from training future models on publisher-owned lyrics. For now, the agreement is a step towards addressing copyright concerns in the AI ecosystem.