Anthropic, an Amazon-backed AI firm, has settled in part of a copyright infringement lawsuit filed by major music publishers.
The agreement, approved by U.S. District Judge Eumi Lee, aims to address concerns about Anthropic's alleged misuse of copyrighted song lyrics in training its AI model, Claude.
In October 2023, Universal Music Group, Concord Music Group, ABKCO, and other publishers accused Anthropic of training its AI system lyrics from over 500 copyrighted songs without permission.
The lawsuit identified instances where Claude copied significant chunks of lyrics from popular songs like Beyoncé's "Halo" and Maroon 5's "Moves Like Jagger." Unlike licensed services such as Genius, the publishers claimed that Anthropic didn't pay any fees for using the material.
They claimed that Anthropic erased copyright management information from the data it used, further violating copyright protections.
Under the interim agreement, Anthropic will maintain and apply its existing copyright protection measures in current and future AI models.
The agreement also includes a provision for music publishers to intervene if the measures fail. If an issue arises, publishers can inform Anthropic in writing, allowing the company to investigate and address it. The court is still ready to handle any unresolved matters.
Anthropic defended its actions, claiming that Claude did not intend to infringe copyright and noting existing measures to avoid such exploitation.
The company reiterated that including potentially copyrighted information in generative AI training constitutes fair use under copyright law.
The settlement constitutes a significant step forward in developing convergence between copyright law and artificial intelligence.
While the music publishers continue to seek preliminary injunctions to prevent Anthropic from using their lyrics for training, this arrangement highlights the more significant issues of balancing innovation and intellectual property rights in AI development.