Music Publishers Say Anthropic Likely Used Pirated Songbooks of Ariana Grande & More for AI Training

The group of music publishers bringing a landmark copyright infringement lawsuit against Anthropic say they’ve recently learned that the artificial intelligence giant illegally downloaded sheet music from online pirate libraries for its training models.

The publishers — including Universal Music Group (UMG), Concord and ABKCO — raised the issue in a Monday (Aug. 11) motion to amend their 2023 lawsuit against Anthropic. The case centers on claims that Anthropic trained its large language model Claude on copyrighted song lyrics, and that the chatbot now spits out copycat lyrics when prompted.

Now, UMG and the other publishers say they’ve learned about another area of alleged wrongdoing by Anthropic: At least seven million books in Anthropic’s training arsenal were illegally torrented from the pirate libraries Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi).

The publishers believe their copyrighted lyrics are among these 7 million pirated books, since LibGen’s library contains thousands of illegal copies of sheet music and songbooks. This includes the sheet music to Elton John’s “Tiny Dancer” and Ariana Grande’s “7 Rings,” as well as music and lyric compendiums with names like The Legendary Songs of Don McLean, The Beach Boys – Greatest Hits, Adele – The Complete Collection and Katy Perry Songbook.

“Even a cursory review of the books that Anthropic torrented from LibGen and other pirate libraries reveals that many, many of these books contain publishers’ works,” reads the motion. “Yet, inexplicably, Anthropic never disclosed to publishers in this case that it had used BitTorrent to copy books containing their works from pirate sites in this manner.”

The music publishers say they learned about Anthropic’s torrenting not from the company, but from a recent ruling by Judge William Alsup in a different copyright lawsuit that accuses Anthropic of training Claude on authors’ literary works.

Notably, Judge Alsup ruled in that case that while the so-called “fair use” defense allows Anthropic to use copyrighted material for training, downloading and storing pirated books is very much illegal. The judge held Anthropic liable for the piracy claims and teed up a trial to determine how much money the company owes authors in damages.

UMG and the other music publishers seem to have taken this ruling to heart, as they’re now seeking to bolster their own case against Anthropic with similar allegations of piracy.

“To be clear, publishers wish to move this case forward more quickly than it is currently proceeding, and are frustrated by Anthropic’s slow pace of discovery and failure to disclose this critical evidence,” says the motion. “Because of Anthropic’s failure to disclose its massive torrenting activity, publishers must now start afresh on this issue.”

Discovery in the publishers’ case against Anthropic is currently scheduled to run through January 2026, with a trial theoretically on track to begin next summer.

In a statement shared with Billboard on Wednesday (Aug. 13), the music publishers say they’re amending the lawsuit to “address this illegal mass torrenting of our works” and “hold Anthropic to account for its additional violations of our rights.”

“The extent of Anthropic’s misconduct is astounding, and it goes directly to our infringement claims,” add the publishers. “What’s just as astounding is the fact that Anthropic concealed these facts for over a year of discovery in our case, and this information came to light only through a decision by a federal judge in another case against Anthropic.”

Reps for Anthropic did not immediately return a request for comment.