Sony Music Publishing and Warner Chappell Music have filed a 48-page lawsuit against Anthropic, naming CEO Dario Amodei and co-founder Benjamin Mann personally as defendants. The complaint, filed August 28 in the U.S. District Court for the Northern District of California, was joined by dozens of affiliated publishing entities — 24 tied to Sony and 11 tied to Warner Chappell, according to reporting from XenoSpectrum.
The core allegation: Anthropic allegedly obtained tens of thousands of copyrighted songs, lyrics, and sheet music through pirate libraries and web scraping, then used that material to train and operate its Claude AI models.
What the Publishers Are Actually Claiming
According to the complaint as reported by Mexico Business News, Anthropic allegedly acquired lyrics and sheet music through torrent downloads, scraping, and datasets that included Common Crawl, along with material pulled from licensed lyrics sites such as Musixmatch and LyricFind. The filing reportedly breaks the case into four legal counts, per XenoSpectrum’s review of the document: direct infringement via BitTorrent, contributory infringement against Amodei and Mann individually, direct infringement spanning the full pipeline from scraping to model output, and a claim involving the removal or alteration of copyright management information — details like a song’s title, writer, or rights holder that are typically embedded in a licensed copy.
The publishers also allege that Claude can reproduce their lyrics verbatim when prompted, and that safety guardrails Anthropic added following earlier litigation from UMG, Concord, and ABKCO can reportedly be bypassed just by re-prompting the model, according to Music Business Worldwide’s coverage of the filing.
The Familiar Paper Trail
Much of the alleged sourcing traces back to material that surfaced during a separate, earlier lawsuit brought by authors — the case that resulted in Anthropic’s $1.5 billion settlement. XenoSpectrum reports the new complaint points to roughly 5 million books from Library Genesis and about 2 million from Pirate Library Mirror as a starting point, the same archives at the center of that earlier author dispute.
Music Business Worldwide’s report also cites a quote pulled from that prior litigation, attributed to Mann, describing downloading pirated books “to avoid the trouble of paying for them.” The same report references a 2024 internal Anthropic planning document, reportedly stating “we don’t want it to be known that we are working on this” in connection with a book-scanning operation. These are allegations drawn from the plaintiffs’ filing and prior case records — not findings a court has made in this case.
How Big Is the Number on the Table
The publishers are seeking statutory damages of up to $150,000 per composition for willful infringement, which is the legal ceiling for that category of claim rather than a figure any court has actually awarded here. With “tens of thousands” of compositions named in the complaint, per Music Business Worldwide, the theoretical exposure runs into the billions.
For context, The Next Web’s coverage notes that authors in Anthropic’s earlier settlement received roughly $3,000 per title — split between author and publisher, landing closer to $1,500 per side after the split. That gap between an opening statutory demand and what a prior case actually settled for is typically where these disputes get negotiated, though nothing here suggests this case is headed toward settlement yet.
Not the First, and Probably Not the Last
This lawsuit doesn’t introduce a new legal theory to the AI-and-music dispute — MusicTimes reports that independent publisher Round Hill Music filed a similar suit against Anthropic earlier in August, alleging unauthorized use of hundreds of songs. What’s different here is scale and standing: with Sony and Warner Chappell now both suing, publishing arms connected to all three major music groups — Universal, Sony, and Warner — are actively litigating against Anthropic over the same underlying issue.
There’s also a European precedent hanging over the case. The Next Web notes that a Munich court ruled against OpenAI in November 2025 in a similar lyrics dispute, finding that a model memorizing lyrics counts as reproduction and isn’t shielded by text-and-data-mining exceptions under German law. That ruling doesn’t bind a U.S. court, but it signals how at least one jurisdiction has already approached the same underlying question.
What the Publishers Are Asking the Court to Do
Beyond monetary damages, MLQ News reports the publishers are seeking a permanent injunction against further infringement, a full accounting of Anthropic’s training data and methods, and court-supervised destruction of any infringing copies still in the company’s possession. They’ve also requested a jury trial.
Anthropic hasn’t filed a public response to this specific complaint yet, and no factual findings have been made by the court — everything above reflects the publishers’ allegations as stated in their filing. Worth noting: Anthropic has had mixed results in related litigation before, including a partial win earlier in a separate lyrics case where a court found the company had already conceded certain points by agreeing to guardrail requirements. How this particular case unfolds — and whether it heads toward the kind of large settlement seen in the authors’ case — will likely shape how other rights holders approach similar claims against AI companies going forward.