Sony Music And Warner Music Sue Anthropic For “Largest IP Theft In History”

Sony Music Publishing and Warner Chappell Music, along with affiliated catalog companies including EMI, Hipgnosis, and Jobete/Motown, filed a copyright infringement lawsuit against Anthropic on August 28 in the Northern District of California, accusing the AI company of 'one of the largest and most blatant ongoing thefts of intellectual property in history' for illegally torrenting, scraping, and downloading copyrighted songs to train its Claude models. The complaint names Anthropic CEO Dario Amodei and co-founder Benjamin Mann as defendants, citing the $1.5 billion settlement in the Bartz v. Anthropic case and alleging Mann personally torrented at least five million books from LibGen in June 2021 and employees downloaded two million more from PiLiMi in July 2022. The publishers also accuse Anthropic of scraping lyrics from licensed sites like MusixMatch and LyricFind, using datasets like Common Crawl and The Pile, and running a 'destructive scanning' operation that digitizes and discards physical songbooks.

As Anthropic is heading to its IPO, it’s having fo contend with some big-name lawsuits. Sony Music Publishing and Warner Chappell Music, along with a long list of affiliated catalog companies including EMI, Hipgnosis, and the Jobete/Motown publishing arm, filed a copyright infringement complaint against Anthropic on August 28 in the Northern District of California. The complaint doesn’t just name the company. It names Anthropic CEO Dario Amodei and co-founder Benjamin Mann personally as defendants, accusing them of directing and approving the alleged infringement themselves. The opening line of the complaint sets the tone for the rest of the document: the publishers say they’re suing “to hold accountable the culprits behind one of the largest and most blatant ongoing thefts of intellectual property in history.” They accuse Anthropic of “a brazen campaign of illegally torrenting, scraping, and downloading copyrighted works on a massive scale” to build and profit from its Claude models. Among the songs the filing lists as caught up in this are “Ain’t No Mountain High Enough,” “All I Want for Christmas Is You,” “Eye of the Tiger,” “Here Comes Santa Claus,” and “Paper Rings,” alongside older catalog staples like “Livin’ on a Prayer,” “Great Balls of Fire,” and “Hallelujah.” The torrenting allegations, again Much of the complaint leans on findings from an entirely different case: Bartz v. Anthropic, the author class action that resulted in Anthropic agreeing to pay https://officechai.com/ai/amid-complaints-about-distillation-anthropic-fined-1-5-billion-for-using-books-to-train-its-ai-models/ $1.5 billion to settle claims that it illegally downloaded more than seven million books from pirate sites LibGen and Pirate Library Mirror PiLiMi . That settlement, reached last year, was widely described as the largest copyright recovery in history, working out to roughly $3,000 per book across around 500,000 titles. Sony and Warner are essentially arguing that the same torrenting spree that got Anthropic into trouble with book authors also swept up their songbooks and sheet music. The complaint alleges Mann personally used BitTorrent in June 2021 to pull down at least five million books from LibGen, and that Anthropic employees torrented another two million from PiLiMi in July 2022, all with Amodei’s knowledge and approval. Quoting from the Bartz record, the publishers cite an internal Anthropic message where Mann described setting up software to “pause torrenting if disk space got ‘exhausted,'” which he called “a cute little libgen babysitter.” Another quoted exchange has Mann sharing a link to PiLiMi with a colleague, who wrote back, “zlibrary my beloved.” The publishers argue this history matters because a court already ruled on the underlying conduct. They cite the Bartz decision’s description of Anthropic’s actions as “straightforward piracy but at massive scale,” and use that finding to argue that Anthropic knew exactly what it was doing when it hoovered up books containing sheet music and lyrics for their catalogs. Beyond torrenting: scraping, scanning, and stripped credits The lawsuit goes further than the piracy claims, though. It accuses Anthropic of scraping lyrics directly from licensed sites like MusixMatch and LyricFind in violation of those sites’ terms of service, pulling lyrics out of datasets like Common Crawl and The Pile, and running a “destructive scanning” operation that involves buying used songbooks, digitizing them, and discarding the physical copies. The filing points to an internal planning document from 2024 on that scanning project that reportedly read, “We don’t want it to be known that we are working on this.” There’s also a copyright management information CMI claim, which is a bit more technical but central to the suit. The publishers say Anthropic strips out copyright notices, songwriter credits, and titles when it cleans training data, using extraction tools the complaint says Anthropic’s own founders favored specifically because they were effective at removing that kind of “boilerplate.” The publishers argue this isn’t just tidying up messy text. They say it’s a deliberate move to make it harder to trace infringement back to the source material once it’s baked into a model. Individual liability for Amodei and Mann What stands out here is that Amodei and Mann are named as individuals, not just as executives of Anthropic. The complaint seeks to hold them personally liable for direct and contributory infringement tied to the torrenting, citing testimony from the Bartz case in which Mann reportedly declined to say under oath whether the torrenting was “exactly lawful or not.” The publishers are asking for statutory damages of up to $150,000 per infringed work, an accounting of Anthropic’s training data and methods, and a court order requiring Anthropic to destroy any infringing copies of the catalogs it holds. The IPO backdrop The complaint doesn’t shy away from pointing out Anthropic’s scale. It cites reporting that Anthropic is eyeing a roughly $2 trillion valuation ahead of a planned October IPO, and argues that the $1.5 billion Bartz settlement barely registers against numbers like that. The publishers write that Anthropic “clearly considers that to be just the cost of doing business,” a framing meant to suggest that without bigger penalties, there’s little reason for the model to change. This isn’t the first music industry suit Anthropic has faced. Concord Music Group and Universal Music Group, working off evidence surfaced during Bartz discovery, filed their own suit against Anthropic earlier this year seeking roughly $3 billion over more than 20,000 works. That case and this new one from Sony and Warner run on separate tracks, but they share a legal team, a court, and a basic argument: that Anthropic’s approach to acquiring training data ran through pirate libraries and scraped websites rather than licensing deals. Anthropic has not yet issued a public response to this filing. In prior music publisher litigation, the company has argued that lyrics displayed openly online don’t carry the licensing restrictions publishers claim, and it has generally maintained that training AI models on copyrighted text qualifies as fair use, a position that got partial backing in the Bartz ruling itself, at least as it applies to training rather than acquisition. Whether that argument holds up against a complaint this detailed, backed by internal Slack messages and sworn testimony from a related case, will now be up to a federal judge in San Jose.