Anthropic has entered into a deal to settle part of a copyright infringement lawsuit filed against the maker of the Claude AI model for allegedly distributing protected song lyrics. The agreement, signed Thursday by U.S. District Judge Yumi Lee, requires Anthropic to apply existing guardrails to the training of future AI models and for music publishers to intervene in cases of suspected copyright infringement. It called for establishing procedures.
In October 2023, multiple music publishers including Universal Music Group, ABKCO, Concord Music Group, and Greg Nelson Music were arrested by Anthropic for allegedly training its AI systems based on the lyrics of at least 500 protected songs. filed a copyright infringement lawsuit against. According to the complaint, when Claude was asked for lyrics to songs like Beyoncé’s “Halo,” Mark Ronson’s “Uptown Funk,” and Maroon 5’s “Moves Like Jagger,” the chatbot answered, “Those… provided a response that included all or a significant portion of the lyrics.” lyrics. “
While music publishers acknowledged that platforms like Genius already distribute lyrics online, they noted that these sites pay licensing fees to use protected works. According to the complaint, when Anthropic ingested data found on these sites to train its AI models, it “intentionally removed or altered copyright management information” from the affected songs.
Under the agreement Anthropic signed Thursday, the AI company says it will maintain guardrails it already has in place aimed at preventing its AI models from infringing on copyrighted content. Anthropic plans to apply existing guardrails to future AI systems it develops. The music publisher and Anthropic will work together “in good faith” to resolve any cases in which the guardrails are deemed ineffective, and the court is prepared to resolve any disputes.
Claude is “not designed to be used for copyright infringement, and we have numerous processes in place designed to prevent such infringement,” Antropic said in a statement to The Hollywood Reporter. said in a statement. “Our decision to enter into this provision is consistent with these priorities. We will use AI models that produce potentially copyrightable material in accordance with existing copyright law. We look forward to continuing to demonstrate that its use in training is exemplary fair use.”
The music publishers behind the original complaint are seeking a preliminary injunction barring Anthropic from training future models based on protected song lyrics, and the court is expected to rule in the coming months. It is planned to be lowered.
