Major Music Publishers File Lawsuit Against AI Firm Anthropic for Alleged Copyright Violations

In a groundbreaking legal move, three major music publishers, Universal Music Publishing Group, Concord, and ABKCO, have initiated a lawsuit against the artificial intelligence company Anthropic. This case, reported by Rolling Stone and further confirmed by documents accessed by Pitchfork, has been filed in a federal court in Tennessee, targeting the San Francisco-based tech firm.

The Core of the Conflict: Copyright Infringement Accusations

At the heart of this legal dispute is Anthropic’s AI assistant, Claude. This large language model (LLM), bearing similarities to OpenAI's renowned ChatGPT, is accused of infringing on the music publishers' copyrights. The allegation is that Claude was trained on the publishers' songs, integrating their lyrics into its generated responses without proper licensing. Furthermore, the lawsuit accuses Anthropic of omitting copyright management information, a violation of the Copyright Act of 1976.

Specifics of the Allegations: A List of Iconic Songs

The lawsuit specifically references 500 copyrighted works owned by the plaintiffs. This list includes legendary songs such as Sam Cooke’s “A Change Is Gonna Come,” the Police’s “Every Breath You Take,” and Beyoncé’s “Halo.” The plaintiffs describe the infringement as systematic and widespread, holding Anthropic accountable not only for Claude’s actions but also for the infringements committed by its users.

Anthropic's Background and Controversial Funding

Anthropic, established by former OpenAI employees in 2021, has attracted funding from notable entities like Google and Zoom. In a significant funding round in April 2022, the company secured $500 million led by Sam Bankman-Fried, the founder of the now-defunct cryptocurrency exchange FTX. This investment is now under scrutiny, as the Department of Justice alleges it originated from FTX customer funds. Recently, Anthropic announced a significant investment from Amazon, amounting to “up to $4 billion,” granting the e-commerce giant a minority stake.

The Challenge of Proving Copyright Infringement

A critical challenge for the plaintiffs is the proprietary nature of commercial LLMs' training sets. The specifics of how or where Anthropic might have sourced the copyrighted material are unknown, given the secretive nature of these datasets. This claim is inferred largely from Claude's output of copyrighted material. Anthropic, in its documentation, states that Claude's models are trained on a mix of publicly available internet data, licensed datasets, and user-shared data.

Examples of Alleged Infringement: Lyrics and New Songs

The lawsuit highlights instances where Claude produced lyrics from songs like Katy Perry’s “Roar,” Gloria Gaynor’s “I Will Survive,” and others when prompted. It also points to occasions where Claude generated new song lyrics that incorporated existing copyrighted material.

Implications and Requests from the Lawsuit

The outcome of this lawsuit could have far-reaching implications for copyright law in relation to AI tools. While the U.S. Copyright Office has issued guidelines about AI-generated works, the legality of training LLMs on copyrighted material remains unclear. The publishers are seeking various remedies, including a permanent injunction against infringement, damages, and legal fees. They also demand that Anthropic disclose the methods used for data collection and training, as well as the destruction of all infringing copies under court supervision.

Conclusion: A Landmark Case in AI and Copyright Law

This lawsuit represents a significant moment in the intersection of AI technology and copyright law. The resolution could set a precedent that may redefine how AI companies interact with copyrighted material, impacting the future of artificial intelligence and its applications in various industries.