Major music publishers are asking a federal judge in San Jose to call it early and rule that Anthropic illegally copied song lyrics to build its Claude chatbot, shutting down the company’s attempt to hide behind a fair use defense. The filing tees up a high-stakes, early test of whether training AI on copyrighted lyrics can be shielded by fair use or treated as straight-up infringement.
Publishers Move For A Pretrial Ruling
According to Reuters, the publishers are urging U.S. District Judge Eumi K. Lee to decide before trial that Anthropic infringed their copyrights and to throw out the company’s fair-use argument. Their motion lays out the legal roadmap, naming lead counsel for the publishers and identifying Sonal Mehta of Wilmer Cutler Pickering Hale & Dorr as Anthropic’s representative counsel in the case.
Case Details And The Claims
The suit, Concord Music Group Inc. v. Anthropic PBC (No. 5:24-cv-03811), accuses Claude of generating verbatim or near-verbatim versions of song lyrics and cites roughly 500 example works in the complaint. Court filings detail the publishers’ claim that Claude’s training data and outputs have siphoned off licensing revenue that should flow to songwriters and rights holders. Justia records the case number and the filings that underpin the motion.
Where This Fits In The Wider AI Copyright Fight
Judges around the Bay Area are already divided on how fair use applies to AI training. In a prominent June 2025 decision, one court described training on legitimately purchased books as "quintessentially transformative" while still leaving room for liability if pirated copies are retained. White & Case breaks down how those rulings drew sharply different lines on market harm and transformation. Anthropic later agreed to a proposed $1.5 billion settlement with authors in 2025, a deal detailed by the Authors Guild.
Publishers' Market Harm Argument
The publishers argue that Claude’s AI-generated lyrics function as derivative or substitute works that go head-to-head with licensed lyrics and erode the markets for songwriting royalties. Anthropic has denied the allegations, and the company did not immediately respond to requests for comment on the publishers’ latest filing, according to Reuters.
What’s Next
The publishers are urging Judge Lee to rule before a jury ever hears the case, a move that could yield a decisive opinion on whether outputs from models trained on copyrighted music qualify as fair use. The motion lands as other complaints tied to Anthropic stack up, including a January lawsuit that accuses the company of using tens of thousands of additional works and seeks roughly $3 billion in damages, as reported by TechCrunch.
Whether Judge Lee opts for an early ruling or sends the clash to a jury, the outcome will ripple far beyond one courtroom. Songwriters, streaming platforms, and every company leaning on large language models will be watching closely as San Jose becomes a key battleground in the fight over copyright and AI training data.
