Anthropic asked U.S. District Judge Eumi Lee to find that training Claude on millions of song lyrics was fair use, a question that could shape other AI copyright cases.
Lee heard arguments in San Jose on Oct. 21 without issuing a ruling, Reuters reported. The hearing addressed Anthropic's use of lyrics in training and the publishers' claim that Claude can produce competing work.
Universal Music Group, Concord and ABKCO brought the case in 2023. Their complaint alleges infringement involving lyrics from hundreds of songs, including works by Beyonce, the Rolling Stones and the Beach Boys.
The publishers say Claude can generate lyrics that compete with human songwriters' work. Their attorney, Matt Oppenheim, said the chatbot received nearly 120,000 requests for synthetic lyrics over six months. He argued that AI-generated material could crowd the market.
Lee asked the publishers to explain how Claude's lyric-writing ability supports their claim that demand for their works is threatened. The question goes to the harm the publishers say results from Claude's output, separate from their allegation about the lyrics used in training.
Anthropic attorney Sonal Mehta described the training as transformative and said it leads to new content. She argued that the publishers had identified no specific market harm involving Claude or another large language model. Anthropic says the training helps Claude understand language for work in science, business and education.
In a 2025 ruling on Anthropic's book library, now-retired U.S. District Judge William Alsup found fair use in training on books but rejected that defense for pirated copies retained in a central library. Two days later, U.S. District Judge Vince Chhabria ruled for Meta in another training case while warning that AI training would not be fair use in many circumstances.
Lee has not yet ruled on Anthropic's request.
