Sony Music Publishing and Warner Chappell Sue Anthropic for Up to $150,000 Per Song

My take: For those of us building with AI, this case is not background noise: it is the moment courts start defining whether the models we use every day were trained legally.

Sony Music Publishing and Warner Chappell filed a 48-page complaint on August 29 in California against Anthropic, CEO Dario Amodei, and co-founder Benjamin Mann, named personally. They allege Claude was trained on "tens of thousands" of copyrighted song lyrics and sheet music downloaded without authorization from torrent sites, Common Crawl, and Library Genesis. Statutory damages sought reach up to $150,000 per infringed work and $25,000 for each instance of copyright management information stripped from text. It is not the only active lawsuit: in January 2026, Universal Music, Concord, and ABKCO filed an action seeking more than $3 billion for over 20,000 works.

What matters here is not the size of the claim. What matters is the precedent. If courts rule that training on copyrighted content required prior licenses, every current language model is potentially exposed, and so are the products built on top of them.

If you rely on AI models to generate content in your business, do you know where the training data for that model came from?