Sony And Warner Sue Anthropic Over Same Piracy

Sony And Warner Filed In California On August 28

Sony Music Publishing and Warner Chappell Music filed suit against Anthropic, chief executive Dario Amodei, and co-founder Benjamin Mann in the US District Court for the Northern District of California late on August 28, 2026, according to TechCrunch and Music Business Worldwide, both of which reviewed the complaint. The filing accuses Anthropic of running what the publishers call "one of the largest and most blatant ongoing thefts of intellectual property in history," seeking damages for what the complaint frames as a deliberate, years-long harvesting operation rather than an isolated lapse.

The complaint traces two acquisition methods. It alleges Mann personally downloaded more than five million pirated books via BitTorrent from the shadow library Library Genesis in June 2021, then alleges Anthropic employees downloaded a further two million books from Pirate Library Mirror in July 2022. Both shadow libraries carry sheet music and song lyrics bundled inside book files, which the publishers say is how Claude's training data absorbed their catalogue without a license. Separately, the suit alleges Anthropic scraped lyrics directly from the licensed lyrics sites MusixMatch and LyricFind, and drew further material from the Common Crawl, The Pile, and Books3 datasets.

The Same Facts Already Cost Anthropic 1.5 Billion Dollars

Anthropic settled a nearly identical claim in September 2025, when a federal judge in the Bartz case ordered the company to pay authors 1.5 billion dollars over the same Library Genesis and Pirate Library Mirror downloads now cited against it by the music publishers. That earlier ruling drew a specific line: training an AI model on copyrighted text was lawful, but acquiring the training copies through piracy was not. Sony and Warner Chappell are not asking a court to redraw that line. They are asking it to apply the line a judge already drew, to a different set of copyrighted works.

At 150,000 dollars per willfully infringed work, a plaintiff group only needs a few thousand confirmed songs to reach billion-dollar exposure before a single copyright-management-information penalty is added. The publishers' own filing names songs including "Ain't No Mountain High Enough," "Eye of the Tiger," "Livin' On a Prayer," and "Uptown Funk" as examples pulled from the training data, a level of specificity that suggests the identification work behind the "tens of thousands" figure is already largely done.

What This Changes For Anyone Buying Claude In Europe

European companies licensing Claude for content, marketing, or internal knowledge tools now have a second live rightsholder class litigating the exact same training-data provenance question, with the underlying facts largely pre-established by a judge in a different case. That combination, admitted conduct plus a fresh set of statutory-damages plaintiffs, is what actually changes here: it is not a new legal theory, it is a new invoice on an old admission. A rightsholder class does not need to prove Anthropic pirated content from scratch; it needs only to show its works were inside the same already-conceded pirated corpus, which is a documents question, not a liability fight.

The practical consequence for a European buyer is procurement, not litigation risk. Any company that has not asked its AI vendor, in writing, which training corpora and which acquisition method underlie the model it is paying for, is relying on an assumption that the two cases above show is no longer safe. Music publishers won the identification fight by cross-referencing Anthropic's own admitted datasets against catalogue metadata; any rightsholder with a comparable catalogue, stock photo libraries, European news publishers, technical documentation vendors, can run the same cross-reference against the same admitted list of datasets. The Sony and Warner Chappell case is not the last filing built on the Bartz admission. It is a template for the next one.

We do this for everyone trying to keep up with what technology is doing to our lives. The people who build it, and the people it happens to. The Servola Journal exists so that what we learn belongs to all of them.

Nobody pays us for this. No ads, no paywall, free to everyone. We just believe that understanding what's happening to all of us shouldn't depend on who can afford to pay for it.

If it gave you something today, tell us to keep going. Follow us, leave a like, or write a positive comment. We read every one, and they are what keeps us going.

Read next: What Killed Mechanical Turk Was Not Competition  |  The AI Cyber Letter Is Also A Liability Record