The Allegation Targets Acquisition, Not Just Training
That distinction matters, because it lands exactly on the line drawn by earlier rulings. The complaint characterizes the conduct as large-scale illegal torrenting, scraping and downloading, naming piracy archives including Library Genesis and Pirate Library Mirror as sources of books containing lyrics and sheet music — the plaintiffs cite examples including "Livin' On a Prayer," "Hallelujah" and "Great Balls of Fire." Beyond acquisition, the complaint lists separate allegations: further distribution via BitTorrent, scraping licensed lyrics sites, digitizing physical songbooks, stripping copyright management information, and training Claude in ways that encouraged memorization of lyrics. That last category carries its own damages measure — up to $25,000 for each alleged removal of copyright management information, calculated separately from the up to $150,000 per willfully infringed work. The publishers also claim Anthropic used their lyrics to teach Claude to generate large volumes of "new" AI lyrics functioning as harmful market substitutes.
Why the Acquisition Question Carries So Much Weight
The reference point is Anthropic's earlier $1.5 billion settlement with authors, in a case where a judge found the company had downloaded more than seven million pirated books. Earlier, in June 2025, Judge William Alsup concluded that training on books was "exceedingly transformative" and fair use — while reaching the opposite conclusion about building a permanent library from pirated copies. In other words, US courts have split this into two questions: whether training is fair use, and how the training material was obtained. The first has gone relatively well for AI companies; the second has not. This complaint clearly aims at the second. One international reference point: a Munich court ruled in November 2025 that memorizing lyrics inside a model constitutes reproduction, and that the text-and-data-mining exception does not cover it.
How Many Suits Is This Now
Anthropic now faces litigation from all three major publishing groups: Universal Music Publishing, Concord and ABKCO sued in Nashville in October 2023 over roughly 500 songs; the same publishers filed again in January 2026 covering more than 20,000 works and seeking over $3 billion; BMG sued in March 2026 over 493 compositions; Round Hill Music filed on August 17; and now Sony and Warner Chappell. The practical implication for readers is limited but worth tracking: how these cases resolve will shape what corpora model vendors can use, and how much compliance room products generating lyric-like content have. The damages figure, though, is a theoretical maximum derived from statutory caps — the actual outcome depends on how many works are found infringed, whether infringement is deemed willful, and whether it settles. Treating "billions" as a conclusion is premature.