Anthropic Seeks Fair-Use Win in Song Lyrics Case

A federal judge in California heard arguments on 21 October over whether Anthropic was entitled to train its Claude chatbot on song lyrics, and ended the day without a ruling. Anthropic asked US District Judge Eumi Lee to find that its use of “millions of song lyrics” is protected as fair use under US copyright law, and so throw out a lawsuit brought by three music companies, Reuters reported via The Straits Times.

The stakes go beyond this case. Lee pressed both sides “to help her determine whether it will become the first AI copyright case to reach trial”, Bloomberg Law reported. She came, it noted, “armed with eight binders of case materials”.

The argument over harm

Much of the questioning centred on market harm, whether Anthropic’s copying hurts the market for, or value of, the copyrighted lyrics. It is one of four factors US courts weigh in deciding fair use, Bloomberg Law noted. According to Reuters, Lee also questioned the music companies’ claim that Claude’s ability to write lyrics threatens their market.

For Anthropic, Sonal Mehta of WilmerHale argued that its training leads to new, transformative content, which would put it squarely under fair use. She said AI had not reduced demand for lyrics, that the plaintiffs could not point to any specific instance of market harm from Claude or other large language models, and that Universal Music Group itself has called AI “overwhelmingly a net positive” for the music industry.

For the music companies, Matt Oppenheim of Oppenheim + Zebrak said Anthropic misuses their lyrics to generate competing content that harms the people who write songs. Large language models such as Claude “could overwhelm the market with AI-generated content very easily”, he said.

Anthropic has argued in the case that its training makes transformative use of lyrics to “help Claude understand human language and enable progress and productivity in science, business and education”.

How the case got here

Universal Music Group, Concord and ABKCO sued Anthropic in 2023, alleging that using their lyrics to train Claude to respond to prompts infringed their copyrights in hundreds of songs, by artists including Beyoncé, the Rolling Stones and the Beach Boys, Reuters reported.

The question Lee is weighing has already split other judges in Northern California. In 2025, Judge William Alsup sided with Anthropic in a case over books, finding that training its AI on them was fair use. Two days later, Judge Vince Chhabria ruled for Meta in a similar lawsuit but warned that AI training “in many circumstances” would not be fair use.

What happens next

There is no ruling yet, and no date for one. If Lee grants Anthropic’s request, the case ends without a trial. If she does not, it could become the first AI copyright case to reach trial, the question Bloomberg Law says she spent the hearing testing.

The decision matters well beyond music. Fair use was the deciding question in both 2025 book cases, and those rulings show judges can weigh it differently. A third view from Lee, in a case built on lyrics rather than books, would add to that picture.