Image: The Verge / Shutterstock
Anthropic has made a deal to settle parts of a copyright infringement lawsuit brought against the maker of the Claude AI model for allegedly distributing protected song lyrics. The agreement was signed off by US District Judge Eumi Lee on Thursday, requiring Anthropic to apply existing guardrails in the training of future AI models and to establish a procedure for music publishers to intervene when copyright infringement is suspected.
In October 2023, several music publishers including Universal Music Group, ABKCO, Concord Music Group, and Greg Nelson Music filed a copyright infringement lawsuit against Anthropic for allegedly training its AI system on lyrics from at least 500 protected songs. According to the complaint, when Claude was prompted for the lyrics to songs like Beyoncé’s “Halo,” Mark Ronson’s “Uptown Funk,” and “Moves like Jagger” by Maroon 5, the chatbot provided responses “that contain all or significant portions of those lyrics.”
While the music publishers acknowledged that platforms like Genius already distribute lyrics online, they noted that those sites pay a licensing fee to use the protected works. Anthropic “intentionally removed or altered copyright management information” from the impacted songs when ingesting data found on those sites to train its AI models, according to the lawsuit filing.
Under the agreement signed by Anthropic on Thursday, the AI company says it will maintain the guardrails it has already implemented that aim to prevent its AI models from infringing on copyrighted content. Anthropic will also apply its existing guardrails to any future AI systems it develops. Music publishers and Anthropic will work together in “good faith” to resolve any instances where the guardrails are deemed ineffective, with the court ready to settle any disputes.
Claude “isn’t designed to be used for copyright infringement, and we have numerous processes in place designed to prevent such infringement,” Anthropic said in a statement to The Hollywood Reporter. “Our decision to enter into this stipulation is consistent with those priorities. We continue to look forward to showing that, consistent with existing copyright law, using potentially copyrighted material in the training of generative AI models is a quintessential fair use.”
The music publishers behind the original complaint have requested a preliminary injunction to bar Anthropic from training future models on their protected song lyrics, with the court expected to issue a ruling in the coming months.