Music publishers demand billions of dollars from Anthropic for stealing song lyrics

Music publishers have filed a new lawsuit against Anthropic; Axios says the lawyers used extremely strong language, calling the case "one of the largest and most brazen intellectual property thefts in history." The complaint centers on the alleged unauthorized use of song lyrics to train the Claude model.

The suit landed late Friday in the U.S. District Court for the Northern District of California. Defendants named include Anthropic and its co‑founders, Dario Amodei and Benjamin Mann. The plaintiffs are seeking $150,000 per claimed copyright infringement — so, on paper, the dollar figures could balloon into the billions.

This isn’t the company's first clash with music rights holders. Universal Music Group, Concord, and ABKCO sued Anthropic in 2023; BMG filed in March; and independent publisher Round Hill Music submitted a claim just days ago. Frankly, the filing reads like another round in a running tally of industry litigation.

Legally, the complaint is built on two strands. One strand leans on documents unearthed in Bartz v. Anthropic (that case alleged massive ebook torrenting via Library Genesis and Pirate Library Mirror). Anthropic settled that suit for $1.5 billion and agreed to destroy pirate copies used or prepared for model training. Song lyrics, i.e., the things people usually fetch from lyric sites rather than ebook torrents, probably weren’t in those e‑book packs — so plaintiffs seem to be using those disclosures to argue a pattern of behavior rather than to prove a direct pipeline of lyric data.

The other, more concrete strand accuses Anthropic of scraping licensed lyric services (e.g., MusixMatch and LyricFind) and training Claude on that material. The publishers claim Claude can reproduce lyrics nearly verbatim; the complaint cites several songs as examples:

• Ain't No Mountain High Enough
• All I Want for Christmas is You
• Eye of the Tiger
• Here Comes Santa Claus
• Paper Rings

That strategy echoes The New York Times’ approach in its suit vs. OpenAI, Microsoft, and Perplexity — not invoking fair use as a shield but pointing to instances where models spit back substantial, paid content and thereby undercut subscriptions. The NYT case is still pending, and this music-publisher action will likely follow the same slow, document-heavy path: motions, discovery fights, and—probably—appeals. Expect a long, bumpy process rather than a quick resolution.