A coalition of major music publishers, including Universal Music Group, Concord Music Group and ABKCO Music, has filed a new federal lawsuit in the U.S. District Court for the Northern District of California accusing AI company Anthropic of illegally downloading more than 20,000 copyrighted musical works — including sheet music, lyrics and compositions — via torrent and other pirate sources to train its Claude models and otherwise exploit the copyrighted material without authorization, seeking over $3 billion in statutory damages and alleging both copyright infringement and Digital Millennium Copyright Act violations in what could become one of the largest non-class action copyright cases in U.S. history. The complaint claims that Anthropic’s leadership knew of “sketchy” unlicensed data sources yet proceeded with the acquisitions and that the resulting AI output continues to generate works that compete with legitimate rights-holder uses, further damaging the music licensing market and justifying significant statutory penalties.
Sources
https://techcrunch.com/2026/01/29/music-publishers-sue-anthropic-for-3b-over-flagrant-piracy-of-20000-works/
https://www.reuters.com/legal/litigation/anthropic-faces-new-music-publisher-lawsuit-over-alleged-piracy-2026-01-28/
https://ipwatchdog.com/2026/01/29/music-publishers-file-new-piracy-suit-against-anthropic-alleging-mass-torrenting-copyrighted-works/
Key Takeaways
• Music publishing giants are escalating legal pressure, alleging Anthropic pirated more than 20,000 works via torrent/shadow libraries to train and deploy AI models, seeking over $3 billion in damages.
• The complaint alleges Anthropic leadership knowingly used dubious acquisition channels and that Claude models still produce infringing outputs, undermining copyright protections and the proper licensing market.
• This lawsuit may set major legal precedent for how AI developers source and use copyrighted material, with potential implications for statutory damages, data acquisition practices, and future AI regulation.
In-Depth
Major music publishers have just thrown a legal gauntlet at Anthropic, the AI developer behind the Claude family of large language models. Backed by industry heavyweights like Universal Music Group, Concord Music Group and ABKCO, the publishers have filed a sweeping copyright infringement complaint in the U.S. District Court for the Northern District of California that alleges Anthropic illegally acquired more than 20,000 copyrighted musical works — including sheet music, song lyrics and compositions — from pirate sites to fuel the training of its AI models. The damages sought exceed $3 billion under U.S. statutory copyright law, which would make this among the largest non-class action copyright cases on record if the court agrees to award the full amount.
At the heart of the publishers’ allegations is a claim that Anthropic didn’t just use unlicensed material inadvertently or as part of broad dataset ingestion, but that company executives — including founders and senior leadership — knowingly turned to BitTorrent and other so-called shadow libraries to download massive troves of copyrighted content despite internal awareness that these sources were “sketchy” and potentially illegal. According to the complaint, that torrenting activity wasn’t merely incidental: it formed a core part of building the training data that underpins Claude’s generative capabilities, meaning the models can reproduce or approximate copyrighted lyrics and compositions without authorization.
The complaint goes beyond simple copyright infringement counts to include charges under the Digital Millennium Copyright Act (DMCA) for stripping or altering Copyright Management Information (CMI) — essentially removing identifying metadata from the works as they were ingested into training datasets — which the publishers argue was done intentionally to facilitate uncensored output from the AI. That allegation underscores the broader claim that Anthropic’s practices undercut the traditional music licensing market by providing unlicensed access and utility to content that normally generates revenue for songwriters, composers and their publishers.
This lawsuit arrives amid intensifying legal scrutiny of generative AI training practices. Publishers’ lawyers and industry observers have watched developments over the past several years as courts, including a judge in the well-publicized Bartz litigation involving authors and Anthropic, have wrestled with how fair use doctrine applies to AI training. While some rulings have found that training on copyrighted material can qualify as fair use, the twist in the current case is the allegation that the way the material was obtained — through piracy and unlicensed downloads — falls outside any fair use protection and constitutes willful infringement.
If the publishers prevail, the statutory penalties alone could dwarf prior settlements and judgments in AI copyright cases, potentially reshaping how data is sourced for large AI models across the industry. Even if the full damages aren’t awarded, the case may force developers to adopt far more transparent and rights-compliant practices, pushing toward comprehensive licensing frameworks that compensate creators fairly. For Anthropic, a ruling against them could not only result in a multibillion-dollar financial hit but also invite injunctions or mandates on how training datasets are assembled and disclosed, altering the company’s business model.
The filing explicitly ties the alleged conduct to ongoing market harm, arguing that the unauthorized use of copyrighted works devalues licensing deals and diminishes creators’ control over how their intellectual property is used and monetized. That narrative resonates across the broader creative industries, where film studios, book publishers, and visual artists have expressed similar concerns about AI systems trained on unlicensed content. With this lawsuit, music publishers are staking a firm legal claim that protecting copyright in the age of AI isn’t merely about limiting output but about enforcing how data is sourced, curated and compensated.
For now, the lawsuit will proceed through the federal court system, and interest from both legal analysts and AI developers suggests it could become a bellwether case that influences future legislation, licensing norms and the broader relationship between artificial intelligence innovation and intellectual property rights.