Anthropic and several record labels in the USA have reached a legal settlement, marking an initial victory for the music industry. The AI provider has agreed that its AI models will no longer output copyrighted song lyrics or use them to create similar texts. This agreement stems from a lawsuit filed in the fall of 2023. However, the contentious issue of whether music and lyrics can be used for training AI models remains unresolved.
The plaintiffs include Universal Music, Concord Music, and ABKCO, which represents artists like the Rolling Stones. They accused Anthropic of using millions of songs to train its AI models and then reproducing them upon user request. Over 500 songs were selected as representatives for the lawsuit, including works by Katy Perry, Beyoncé, and many other well-known artists.
Anthropic has committed to implementing additional safeguards in future AI models to ensure that the copyrighted songs of the plaintiffs are not used as output by the AI. The agreement also allows record labels to intervene if these safeguards fail to function as intended. They can notify the AI provider of such cases, and the provider must respond immediately.
Anthropic assures that it is unlikely for the AI models, particularly the AI chatbot Claude, to reproduce song lyrics verbatim. There are already protective measures in place to prevent this. Nonetheless, the safeguards will be strengthened.
A court ruling on the use of song lyrics and music for AI training is still pending, which could take several months. The plaintiffs accuse Anthropic of “direct copyright infringement, contributory infringement, vicarious infringement, and removal or alteration of copyright management information in violation of the Copyright Act.” The newly signed agreement explicitly states that this issue remains undecided. “In particular, the parties continue to dispute the publishers’ request for a preliminary injunction, which is neither resolved nor affected by this agreement,” the statement reads.
Anthropic, like other AI providers, has relied on the US principle of Fair Use, which allows the use of copyrighted material if it benefits the public. Besides this lawsuit, there are other lawsuits with similar allegations. For instance, GEMA, the music rights organization, has filed a lawsuit against OpenAI in the Munich Regional Court. Additionally, there are lawsuits from publishers, with the most prominent example being the lawsuit by the New York Times. Smaller publishers and individual artists also seek to protect their interests in court. Lawsuits have also been filed against image generator providers, such as the copyright lawsuit by Sarah Andersen against Stability.
The outcome of these legal proceedings will significantly impact how AI companies use copyrighted material for training their models. The music industry is keen to ensure that their content is not used without proper authorization, and these cases will likely set important precedents for the future.
While the current settlement represents a step forward for the music industry, the ongoing legal battles highlight the complex intersection of AI technology and copyright law. As AI continues to evolve, finding a balance between innovation and intellectual property rights will be crucial.
Case Number: Case 5:24-cv-03811-EKL