Sony Music and Warner sue Anthropic over AI copyright claims

The complaint, Sony Music Publishing (US) LLC et al. v. Anthropic PBC et al., was filed on Aug. 28, 2026, in the U.S. District Court for the Northern District of California, San Jose Division. The case number is 5:26-cv-09217, and the plaintiffs have demanded a jury trial.

According to the 48-page complaint, Anthropic personnel used BitTorrent to obtain at least 5 million books from Library Genesis in June 2021 and at least 2 million more from the Pirate Library Mirror in July 2022. The plaintiffs allege that the material included books containing lyrics or sheet music for compositions such as “Ain’t No Mountain High Enough,” “All I Want for Christmas Is You,” “Eye of the Tiger,” “September” and “Hallelujah.”

The complaint also alleges that Anthropic scraped lyric websites including Musixmatch and LyricFind, used datasets such as Common Crawl, The Pile and Books3, and made unauthorised copies for Claude training and fine-tuning. The publishers further allege that Claude has generated verbatim or near-verbatim lyrics and derivative material in response to prompts. Exhibit B identifies tens of thousands of allegedly affected works, although the filing says its lists are illustrative and non-exhaustive.

Those are allegations advanced by the plaintiffs, not findings by the court. The complaint includes four counts: direct copyright infringement by torrenting against all defendants; contributory infringement against Anthropic co-founders Dario Amodei and Benjamin Mann; direct infringement against Anthropic; and alleged removal or alteration of copyright-management information.

The plaintiffs seek up to $150,000 per work they prove was wilfully infringed, or actual damages and profits. They also request up to $25,000 per alleged copyright-management-information violation, a permanent injunction, an accounting of training data and acquisition methods, supervised destruction of infringing copies, legal fees, costs and interest.

This is bigger than one chatbot producing a suspiciously familiar chorus. The publishers allege unauthorised training copies damaged licensing revenue, weakened lawful licensing markets and reduced catalogue value. They also say AI-generated lyrics compete with the writers and publishers whose work helped make the systems useful.

The case tests whether copying during data collection and training is legally defensible, and whether outputs reproducing protected expression create a separate problem. The plaintiffs say titles, author names and other identifying information were altered or removed, which they claim violates Section 1202 of the U.S. Copyright Act.

Anthropic disputes the publishers’ claims. In a statement reported by TechCrunch, the company said: “We disagree with the publishers’ claims and we intend to defend ourselves robustly in court.” No substantive answer, motion or ruling was publicly identified in the supplied materials.

The complaint’s language is intentionally explosive, but the legal test will be evidence, not adjectives. If the allegations about torrenting, training copies and recognisable lyric outputs are substantiated, “move fast and break things” starts looking less like a startup slogan and more like a licensing strategy with lawyers attached. AI companies need useful data, but “useful” cannot quietly become “free.” Watch this case closely — and do not treat the filing’s allegations as settled truth.