Major music publishers, including Universal Music Group, Concord Music Group, and ABKCO, have reached a pivotal agreement with AI company Anthropic regarding the use of copyrighted song lyrics in training artificial intelligence. This agreement follows a lawsuit filed in 2023, accusing Anthropic of copyright infringement for using lyrics from over 500 songs by artists like Katy Perry, the Rolling Stones, and Beyoncé.
Key Details of the Agreement
Under the deal approved by U.S. District Judge Eumi Lee, Anthropic’s AI chatbot, Claude, will no longer provide copyrighted song lyrics or generate new lyrics based on such material. Additionally, the company has committed to maintaining strict guardrails to prevent copyright violations.
Anthropic stated, “Claude isn’t designed to be used for copyright infringement, and we have numerous processes in place to prevent such infringement. This agreement aligns with our commitment to fair use within existing copyright laws.”
Background of the Lawsuit
Music publishers initially filed the lawsuit in Tennessee federal court, alleging that Anthropic’s use of song lyrics for training its AI undercut existing licensed markets. They cited specific examples, such as Claude providing near-identical lyrics to Katy Perry’s “Roar.”
This lawsuit is the first legal challenge against an AI firm over the use of lyrics in training language models and highlights growing tensions between the creative industries and AI developers.
Terms of the Deal
- Guardrails in Training: Anthropic agreed to implement robust measures to prevent AI-generated outputs from reproducing copyrighted lyrics.
- Publisher Oversight: Music publishers can notify Anthropic if the guardrails fail. The company must investigate and address issues promptly.
- Future Safeguards: The agreement applies to current and future AI training, ensuring continued compliance with copyright law.
Implications for AI and Copyright Law
This case underscores the evolving debate over the use of copyrighted material in AI training. Anthropic maintains that training on copyrighted content falls under “fair use,” but publishers argue that such practices threaten their revenue and licensing models. A ruling on a preliminary injunction, which could further restrict Anthropic’s practices, is expected in the coming months.
Industry Reactions
The agreement is seen as a significant step toward balancing innovation in AI with the rights of creators. It also sets a precedent for how other AI companies might navigate copyright concerns in the future.
The music publishing industry and AI developers are navigating uncharted territory as technology continues to advance. This landmark deal may serve as a framework for future agreements, ensuring that creators are fairly compensated while enabling responsible AI innovation.