Music publishers including Sony, EMI, and Warner Chappell are pressing a copyright case against Anthropic, arguing that the Claude maker’s alleged use of pirate libraries extended beyond books and into copyrighted musical compositions, according to Ars Technica. The lawsuit, filed Friday, alleges that Anthropic’s “illegal torrenting” included “thousands upon thousands” of the publishers’ copyrighted compositions. Ars reports that the publishers claim pirated books Anthropic allegedly plans to keep “forever” for AI training include titles featuring the complete works of the Beatles, a collection of Taylor Swift songs, and “VH1’s 100 Greatest Songs of Rock & Roll.” The complaint also ties the music case to Anthropic’s earlier author settlement. According to Ars, the publishers said Anthropic “got off too light” when it paid authors $1.5 billion after admitting to pirating more than 7 million books to train AI. The publishers argue in the lawsuit that $1.5 billion is not enough to deter a company they describe as having a $2 trillion valuation. The case targets both Anthropic and two of its leaders personally. Ars reports that the complaint alleges Anthropic co-founder Benjamin Mann used BitTorrent beginning in July 2021 to download and upload millions of pirated books from Library Genesis, also known as LibGen. The complaint further alleges that CEO Dario Amodei approved the torrenting; both Mann and Amodei are named individually as defendants, according to Ars. A central part of the publishers’ case is internal messaging cited in the complaint. After the FBI shut down LibGen, the complaint says pirates copied its contents into Z-Library, and that Anthropic later obtained access to a copy through the Pirate Library Mirror, or PiLiMi. Ars reports that Mann allegedly directed employees to torrent PiLiMi shortly after it became available, writing that the mirror dropped “just in time!” A staffer allegedly replied: “zlibrary my beloved.” The publishers use those messages to argue that Anthropic did not merely encounter infringing material incidentally, but relied on torrenting as a fast way to acquire training data. According to Ars, the complaint says metadata in LibGen and PiLiMi catalogs indicated Anthropic torrented at least hundreds of books containing sheet music and lyrics for compositions owned by the publishers. The alleged harm is not limited to copying inputs. Ars reports that the publishers accuse Anthropic of reproducing lyrics verbatim and generating “new” songs that mimic the “heart” of popular works. The complaint gives an example in which Claude could allegedly rely on songbooks or sheet music when responding to a prompt asking it to change Eminem lyrics into a song in “Beyonce’s style.” Those claims remain allegations in an active lawsuit, and the provided material does not include Anthropic’s response. But the filing shows how music-rights holders are adapting book-copyright arguments for AI training into a broader challenge focused on lyrics, sheet music, model outputs, and internal company intent. Who benefits: If the publishers’ arguments gain traction, music publishers and songwriters could gain leverage over AI companies that trained on copyrighted lyrics or sheet music. The immediate benefit is legal leverage, not a proven market outcome. Who's exposed: Anthropic is the primary exposed party, and Ars reports that Dario Amodei and Benjamin Mann are named individually as defendants.