Sony Music Publishing and Warner Chappell Music have filed a major federal lawsuit against Anthropic, accusing the AI company of systematically stealing copyrighted songs to train Claude, the company's generative AI model. The publishers characterize the alleged conduct as "a brazen campaign of illegally torrenting, scraping, and downloading copyrighted works on a massive scale," marking the most aggressive legal challenge the music industry has mounted against an AI developer.
The lawsuit represents a critical inflection point in ongoing tensions between tech companies building large language models and rights holders who control vast catalogs of creative content. Claude has become one of the most widely used AI assistants, competing directly with OpenAI's ChatGPT and Google's Gemini. The question of what training data these systems can legally use remains unresolved in courts, though multiple industries are testing the boundaries through litigation.
Music publishers have emerged as particularly vocal critics of AI training practices. Unlike film studios or book publishers, which have pursued selective licensing deals with some AI companies, the music industry has largely taken an adversarial stance. Warner Chappell, which controls rights to millions of compositions including works by The Beatles, Ed Sheeran, and Stevie Wonder, and Sony Music Publishing, which represents artists like Michael Jackson and Bob Dylan, command enormous leverage. Together they control roughly 20 percent of the global music publishing market.
The complaint appears to center on how Anthropic assembled training datasets for Claude. AI companies typically need vast amounts of text to build functional models. Anthropic has disclosed some training sources but keeps much of its methodology proprietary. The publishers allege the company didn't obtain licenses or seek permission before using copyrighted lyrics, song metadata, or related musical content in its training pipeline. This echoes complaints made against OpenAI by The New York Times, which sued the company in January 2024 for similar alleged copyright infringement at massive scale.
The timing matters. Federal courts have not yet ruled definitively on whether training copyrighted works into AI systems constitutes fair use or copyright infringement. The Second Circuit Court of Appeals is currently considering the Times' case against OpenAI, which could establish precedent. Music publishing lawsuits could accelerate clarity on this question or create conflicting legal precedent, depending on judicial interpretation. Copyright law traditionally protects transformative uses, but courts must decide whether AI training constitutes transformation in the legal sense.
Anthropic has positioned itself as a more safety-conscious AI developer than competitors, emphasizing responsible practices and legal compliance. This lawsuit directly challenges that positioning. The company will likely argue that training on copyrighted material constitutes fair use or that it obtained licensing through third-party datasets. However, if the companies can prove Anthropic deliberately torrented or scraped content rather than licensing it through legitimate channels, the legal exposure becomes substantially higher.
The music industry's aggressive posture contrasts with some other content creators. Taylor Swift and Billie Eilish have publicly criticized AI music generation but haven't sued developers directly. However, the publishing companies' resources and market power enable them to pursue litigation that individual artists cannot. Their case could reshape how AI companies approach music data acquisition going forward, potentially forcing licensing arrangements or pushing companies toward synthetic or licensed training data exclusively.
This lawsuit signals that the music publishing industry intends to fight AI developer practices in federal court rather than negotiate settlements or licensing deals.