UMG / Concord / ABKCO v. Anthropic, $3B AI lyrics copyright suit

Background

In January 2026, music publishers Universal Music Group Publishing, Concord, and ABKCO filed a lawsuit in the Northern District of California against AI company Anthropic, alleging that the training data used to build Anthropic's Claude AI models included the unauthorized wholesale copying of more than 20,000 copyrighted songs, lyrics, sheet music, and musical compositions. The complaint characterized the conduct as "brazen" and "flagrant piracy" and sought $3 billion in statutory damages, which would make it among the largest single non-class-action copyright cases in U.S. history.

Anthropic is expected to invoke a fair-use defense, arguing that training AI on copyrighted material is transformative. In March 2026, the plaintiffs moved for partial summary judgment, contending the evidence of infringement is "overwhelming" and that Anthropic's fair-use theory fails as a matter of law. On March 30, 2026, the RIAA, NMPA, and six allied organizations filed an amicus brief calling Anthropic's unlicensed copying "inexcusable" and urging the court to reject the fair-use argument. No trial date has been set.

Why It Matters for Artists and Fans

Publishing rights, the ownership of song lyrics and compositions, are among the most valuable assets in music, generating income from every use including AI-generated paraphrases and reproductions. If the court finds that LLM training on lyrics constitutes infringement, every major AI company will need licensing deals with publishers before reproducing song text, fundamentally changing the economics of AI development. A ruling in Anthropic's favor would, conversely, signal that all published lyrics in an AI's training corpus are effectively in the public domain for machine-learning purposes, a result songwriters and publishers argue would devastate the economic foundation of composition rights.