Sony Music Publishing and Warner Chappell are taking Anthropic to court, alleging that the company’s Claude AI models have repeatedly reproduced copyrighted song lyrics without authorization. The lawsuit, filed in federal court, accuses Anthropic of systematically ingesting and outputting protected music content, calling the practice a “blatant violation” of U.S. copyright law.
Image Courtesy : cxtoday.com
The music publishers argue that Anthropic trained its models on datasets containing copyrighted lyrics, allowing Claude to generate or closely paraphrase lines from popular songs when prompted. According to the complaint, the companies documented thousands of instances where Claude allegedly produced lyrics from artists across their catalogs — sometimes verbatim, sometimes with minor alterations.
Their position is straightforward: Anthropic is benefiting from creative works it never licensed, and the output proves those works were used during training. The publishers say this isn’t accidental or incidental — it’s structural. They claim Anthropic’s models “memorize and regurgitate” copyrighted content, undermining the licensing market that songwriters and publishers rely on.
Anthropic, for its part, has not publicly commented on the lawsuit, but the company has previously stated that it uses a mix of licensed, public, and synthetic data to train its models. It also emphasizes that Claude is designed to avoid reproducing copyrighted text, though the publishers argue that these safeguards are insufficient and inconsistently applied.
This lawsuit is part of a broader wave of legal action against AI companies. Authors, news organizations, and other rights holders have filed similar complaints, arguing that generative AI systems are built on unlicensed data scraped from the internet. But the music industry’s involvement raises the stakes: lyrics are among the most aggressively protected forms of intellectual property, and publishers have a long history of litigating infringement.
If the court sides with Sony and Warner, the implications could be far‑reaching. AI companies may be forced to license large portions of their training data, purge copyrighted content from existing models, or redesign systems to prevent memorization. For the music industry, a win could establish a precedent that protects their catalogs from being absorbed into AI systems without compensation.
The case also highlights a deeper tension: AI companies argue that training on publicly accessible data is transformative and necessary for innovation, while rights holders insist that access does not equal permission. As generative AI becomes more capable, these conflicts are intensifying — and courts are being asked to define the boundaries of fair use in the age of machine learning.
For now, Sony and Warner are making their position clear. They believe Anthropic crossed a legal line, and they’re prepared to fight it in court.