Writing Team
Sony Music and Warner Music just sued Anthropic, alleging the company torrented tens of thousands of copyrighted songs to train Claude. That's a genuinely good thing for anyone who makes a living creating something ownable, including every marketer building original content for a client.
Key Points
- Sony Music, Warner Music, and other publishers sued Anthropic in federal court, naming CEO Dario Amodei and co-founder Benjamin Mann personally over allegedly torrenting tens of thousands of copyrighted song lyrics and compositions.
- The complaint seeks up to $150,000 per infringed work, plus penalties for removing copyright management information, and calls the alleged conduct "one of the largest and most blatant ongoing thefts of intellectual property in history."
- This follows Anthropic's $1.5 billion settlement with book authors in September 2025, the largest copyright settlement in U.S. history, over the same underlying practice: acquiring training data through pirated sources.
- The case centers on acquisition, not use, meaning the legal exposure comes from how the data was obtained (alleged torrenting and scraping licensed platforms) rather than whether AI training itself is legal.
- A German court already ruled in a related case that AI-generated song lyrics count as unlawful reproduction, putting real judicial weight behind the idea that model operators, not just users, are accountable for what their systems output.
Why This Lawsuit Targets How Anthropic Got the Data
The complaint's real weapon isn't a debate about whether AI training on copyrighted material is fair use. It's a much simpler claim: Anthropic allegedly torrented at least seven million books from pirate libraries LibGen and PiLiMi, according to the-decoder's coverage, and allegedly scraped song lyrics from licensed platforms like MusixMatch and LyricFind in violation of those platforms' terms of service. That's a strategy Anthropic has already lost once. It paid $1.5 billion to book authors in September 2025 for the exact same pattern: not training on copyrighted work itself, but acquiring it through illegal downloads first.
What Makes This Case Different From the Book Settlement
This complaint goes further by naming Amodei and Mann as individual defendants, alleging they personally directed the acquisition of the disputed material. It also targets a more sophisticated dodge: the plaintiffs claim Anthropic trained a commercial Claude model on synthetic data generated by a separate, non-commercial model that had itself learned from the pirated texts, effectively laundering the copyrighted material through an intermediate model. If that allegation holds up, it closes a loophole that has let AI companies claim clean hands on their flagship products while quietly building on infringing foundations elsewhere in the pipeline.
Why Creators and Marketers Should Welcome This Fight
A world where AI companies have to license the content they train on is a better world for anyone whose work is the product. Marketers, writers, musicians, and every business built on original content have a direct stake in whether "we scraped it, so it's ours to use" holds up as a legal defense, because the same logic that lets a lab torrent song lyrics is the logic that would let it strip-mine your brand's content next. A German court already ruled in a related case that AI-generated lyrics count as unlawful reproduction and held the model operator, not the user, responsible for the output. That's the direction the law needs to keep moving. If your team is thinking through how AI tools fit into a growth strategy without exposing your own content to the same kind of uncredited reuse, or you want a second set of eyes on how the AI vendors in your stack actually sourced their training data, our AI marketing services team can help you ask the right questions before you build on top of them.
Anthropic has now lost this exact fight once and is defending it a second time with two of its founders named personally. That's accountability working the way it's supposed to.