Key Music Copyright Trial: Anthropic and Record Labels Debate Fair Use in AI Training
nashnova research
Federal Judge Eumi Lee heard arguments in San Jose on whether Anthropic's use of song lyrics to train Claude counts as fair use — the third federal ruling on the issue, with implications for dozens of pending AI copyright cases.
What is this lawsuit actually about?
Universal Music, Concord, and ABKCO sued Anthropic in 2023, alleging it used lyrics from hundreds of songs without authorization — by artists including Beyoncé, the Rolling Stones, and the Beach Boys — to train its chatbot Claude.
The core dispute is narrow: does using copyrighted lyrics to train an AI qualify as "fair use" — a legal doctrine that permits certain unlicensed uses of protected work?
This means → the court is not deciding *whether* Anthropic used the lyrics — both sides agree it did. The question is whether that use is lawful.
What does each side argue?
Anthropic's central claim: its use of lyrics is "transformative" — the goal is not to reproduce songs but to help Claude understand human language, advancing science, commerce, and education. It cites Supreme Court precedent holding that copyright's primary purpose is to serve the public, not to reward authors.
The publishers counter that Claude can generate large volumes of new lyrics, diluting the market value of their works. They point to evidence that "AI-generated songs are flooding streaming platforms and even charting on Billboard."
In plain terms = one side says "we used lyrics to teach AI to talk, not to sell songs"; the other says "once taught, the AI writes songs that compete with ours — that is not fair."
Why is this called "the third vote"?
Two Northern California federal judges have already ruled on AI-training fair use — and reached opposite conclusions. Retired Judge William Alsup sided with Anthropic, finding that training AI on books qualifies as fair use.
Just two days later, Judge Vince Chhabria ruled in Meta's favor in a similar case, yet warned that AI training "in many instances" does not constitute fair use.
This means → Judge Lee's decision acts as a critical third judicial reference point. If she sides with the publishers, the fair-use defense shifts from 2-1 to 1-2, making it significantly harder for AI companies to invoke in future cases.
How big is the impact on the AI industry?
Writers, news organizations, and visual artists have already filed dozens of high-stakes copyright suits against Anthropic, OpenAI, Meta, and others. This ruling will serve as a major reference for all of them.
This reflects a fundamental legal question facing the AI industry: where is the line for using human-created content to train AI? Courts are still drawing that line, one case at a time.
In plain terms = this is not just a fight over song lyrics — its outcome will signal to the entire industry where the legal red line for AI training actually sits.
市场有风险，内容仅供研究参考，不构成投资建议。