Sony Music and Warner Chappell sue Anthropic, alleging Claude was trained on tens of thousands of pirated songs

Sony Music Publishing and Warner Chappell Music sued Anthropic on August 28, accusing the AI lab of training its Claude models on what the complaint calls tens of thousands of copyrighted songs -- pulled from pirated book archives, licensed lyrics sites, and AI-training datasets -- without ever paying for a license. Filed in the U.S. District Court for the Northern District of California, the suit names Anthropic PBC and co-founders Dario Amodei and Benjamin Mann personally as defendants, the first of five music-industry suits against the company since 2023 to target the founders as individuals rather than just the company. (Naming Amodei and Mann personally raises what's actually at stake for two specific people, not just Anthropic's balance sheet -- personal liability for copyright infringement is unusual, and including it is itself a signal about how the publishers plan to litigate this.)

"Defendants Anthropic and its founders Dario Amodei and Benjamin Mann have conducted a brazen campaign of illegally torrenting, scraping, and downloading copyrighted works on a massive scale in order to develop, operate, and reap enormous profits from Anthropic's 'Claude' series of artificial intelligence models." -- the complaint, paragraph 1

The complaint leans hard on a ruling from a different, earlier case. In 2025, a federal judge in the same courthouse found that Anthropic had illegally torrented more than seven million pirated books from Library Genesis and Pirate Library Mirror to build its training library, calling the conduct "straightforward piracy but at massive scale." That case, Bartz v. Anthropic, ended in a $1.5 billion settlement with book authors -- the largest copyright settlement on record -- which received final court approval on July 20, 2026, five weeks before this new suit was filed. Sony Music Publishing and Warner Chappell say some of the same torrented books contained the lyrics and sheet music to their songs, which is the acquisition claim this suit borrows directly from Bartz.

This is also the fifth music-publisher suit against Anthropic in under three years, and Sony and Warner are late arrivals, not the first movers. Universal Music Group, Concord, and ABKCO opened the sequence in October 2023 over Claude reproducing lyrics like Katy Perry's "Roar" verbatim; they returned in January 2026 with a second suit covering more than 20,000 songs and seeking over $3 billion; BMG followed in March over 493 compositions; Round Hill Music filed August 17. The complaint splits its own claims into two tiers that are worth keeping separate: Exhibit A lists specific songs -- among them "Livin' on a Prayer" and "Great Balls of Fire" -- that the publishers trace directly to the torrented books at the center of the Bartz case, while Exhibit B is far larger: tens of thousands of works the publishers say Anthropic separately harvested by scraping licensed lyrics sites like MusixMatch and LyricFind, "destructively scanning" secondhand books, and pulling from the Common Crawl, The Pile, and Books3 datasets. One exhibit rests on conduct a court has already called piracy; the other is a new, much larger claim the publishers still have to prove.

The complaint also adds a second legal theory beyond piracy: that Anthropic stripped copyright management information (CMI) -- the embedded metadata identifying a song's writer and publisher -- from the text it harvested, a separate violation under the Copyright Act that can carry its own damages of up to $25,000 per work. The publishers allege Anthropic's own internal process for cleaning "garbage" out of training data doubles as a way to erase that attribution trail:

Anthropic disputes all of it. "We disagree with the publishers' claims and we intend to defend ourselves robustly in court," a company spokesperson said in a statement reported by TechCrunch and Music Business Worldwide. The company has one genuine opening the Bartz case left for it: that ruling found the *torrenting* of the books illegal, but left open whether training a model on lawfully-acquired copyrighted text is itself fair use -- a defense several AI labs are actively litigating elsewhere. Sony and Warner's new suit tries to close that gap by treating the acquisition (illegal torrenting, already established in Bartz) and the training use (fine-tuning Claude toward verbatim lyric reproduction, not yet adjudicated) as one continuous, unlicensed act rather than two separate questions.

That fight over training use will not happen in a vacuum. It is the latest round in a pattern now five suits deep, and the pattern itself is worth seeing whole:

The complaint also puts a number on Anthropic's scale that is worth checking against the company's own record: it calls the conduct inexcusable given Anthropic's "staggering $2-trillion-dollar valuation." That figure runs well ahead of what Anthropic has actually disclosed. The company's own most recent funding announcement, its May 2026 Series H, priced Anthropic at $965 billion post-money -- roughly half the complaint's number, and the one Anthropic itself has published.

Anthropic has 21 days from being served to answer or move to dismiss. Whatever it files next, this suit and the pending UMG/Concord/ABKCO case sit in front of courts in the same circuit that just decided Bartz, on the one question Bartz didn't reach: whether training a model on copyrighted lyrics, once you get past how they were acquired, is itself lawful. Whichever way that eventually gets answered, it will apply to every AI lab training on copyrighted text, not just to Anthropic.