Sony Music Publishing, Warner Chappell Music and their affiliates filed a lawsuit against Anthropic and co-founders Dario Amodei and Benjamin Mann, accusing the company of obtaining and copying copyrighted works on a large scale to develop its Claude models. The lawsuit was filed on Friday (28) in the United States District Court for the Northern District of California.

Summary: The lawsuit brings together 35 music publishers and cites tens of thousands of compositions allegedly used without authorization. The plaintiffs seek damages that could reach US$ 150,000 per infringed work, in addition to other remedies provided for under US law.

The petition states that Anthropic allegedly resorted to torrents, automated collection of internet pages and other sources to form datasets used in training Claude. The accusations have not yet been judged, and the company disputes the allegations.

Among the works mentioned in the document 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”. According to the plaintiffs, the set of works listed in the appendices includes tens of thousands of protected compositions.

Publishers point to torrents, scraping and digitized books

One of the main fronts of the lawsuit concerns how Anthropic allegedly obtained part of the material. The plaintiffs state that Benjamin Mann used BitTorrent to download at least 5 million books from Library Genesis, LibGen, in June 2021, with Amodei's authorization.

The petition also points to the download of at least 2 million books from Pirate Library Mirror in July 2022. Part of these files, according to the publishers, contained lyrics and sheet music of compositions administered by the groups involved in the lawsuit.

The plaintiffs also accuse Anthropic of collecting lyrics from licensed websites, including MusixMatch and LyricFind, and of using datasets such as Common Crawl, The Pile and Books3. The company also allegedly digitized physical copies of used books and destroyed the originals after converting them to digital text.

According to the lawsuit, hundreds of songbooks and sheet music collections would have been included in this digitization process. The converted material would become part of an internal library that could later be used in artificial intelligence training.

The publishers also allege that Claude can produce identical or very similar excerpts to the protected lyrics present in the training data. The petition states that mechanisms adopted by Anthropic to prevent the reproduction of protected content could be circumvented with new commands sent to the model.

Lawsuit could generate damages per work

The lawsuit presents four main accusations. They include direct copyright infringement through torrents against the three defendants, contributory infringement attributed to Amodei and Mann, as well as direct infringement and removal or alteration of copyright information attributed to Anthropic.

The plaintiffs ask the court for damages of up to US$ 150,000 per work if the violations are found to be intentional. They also claim up to US$ 25,000 per occurrence related to the alleged removal or alteration of copyright management information.

The publishers also want a permanent injunction to stop the practices considered infringing and require Anthropic to provide information about the data and methods used in training its models, including the identification of protected works incorporated into the training sets.

Anthropic said it disagrees with the accusations and intends to defend itself in court.

More from Radar