Anthropic faces a vast copyright fight. Major music publishers accuse the AI company, its chief executive Dario Amodei, and co-founder Benjamin Mann of copying tens of thousands of songs to train its models. The claims place training data back at the center of the AI industry’s legal problem — and attach a very large price tag.
The lawsuit involves at least 20,000 works, including lyrics and sheet music for songs such as Bon Jovi’s “Livin’ on a Prayer” and Leonard Cohen’s “Hallelujah.” Other songs named as part of Anthropic’s training data include “Ain’t No Mountain High Enough,” “September,” and “Paper Rings.”
Music publishers have also claimed that Anthropic’s models “scraped” smash hits including Mariah Carey’s “All I Want For Christmas Is You.” The named works connect songs associated with Mariah Carey, Marvin Gaye, Tammi Terrell, Survivor, Mark Ronson, and Katy Perry to the dispute, turning an abstract argument about datasets into a catalog of familiar music.
The alleged copying trail
The publishers accuse Anthropic of “illegally torrenting, scraping and downloading” copyrighted works on a massive scale. They claim the company downloaded at least 7 million copies of books from pirate websites, while Mann used BitTorrent to download over five million pirated books and employees downloaded at least two million more books from Pirate Library Mirror.
The music allegations describe several routes into Anthropic’s training material. The company allegedly acquired song copies by scraping lyrics from sites such as Musixmatch and LyricFind, alongside datasets hosted on archive sites such as Common Crawl.
According to the claims, Anthropic then stripped identifying information from songs while processing the texts. That alleged removal denied copyright owners valuable attribution and created a second legal issue beyond the copying itself — because the case also seeks penalties for removing or altering identifying data.
Sony Music Publishing and Warner Chappell are among the music publishers involved. Anthropic has also faced claims from Universal Music Group and Concord Music Group, among others, while BMG and Round Hill Music are named among the music companies connected to the broader dispute.
Billions in possible damages
The financial exposure comes from the scale of the alleged infringement. The publishers are seeking damages that could run into billions of dollars, based on up to $150,000 for each infringed piece of work and $25,000 for each removal or alteration of identifying data.
Applied across at least 20,000 works, those figures produce a legal threat far beyond a routine licensing disagreement. The claim refers to the unlawful copying of “tens of thousands” of copyrighted musical compositions, meaning each song may carry its own separate consequence.
Anthropic has already settled a copyright case against it for $1.5 billion. That earlier settlement does not resolve these music claims, but it gives the current dispute a useful sense of scale: AI training data is no longer a background technical choice when every copied work can become a line item.
Anthropic disputes the publishers’ claims and says, “We disagree with the publishers’ claims and we intend to defend ourselves robustly in court.” Amodei and Mann are named as individual defendants, so the case reaches beyond the company’s balance sheet and into the decisions made by its leadership.
The core question is familiar but unresolved: whether an AI company can gather copyrighted material at industrial scale, process it into training data, and defend the practice after the fact. This lawsuit gives that question a song list, a book count, and a damages formula. The calculator is doing more work than the chatbot.
Based on
