Forty-one million dollars for one dead singer’s catalog is a pricing signal AI companies should read carefully. The owners of Elvis Presley’s music have sued Anthropic, according to Engadget, which cited a New York Times report. The complaint was filed on October 23rd and claims Claude’s maker trained on Elvis songs without permission.
The plaintiffs include Authentic Brands Group, Raleigh Music Publishing, several trusts and at least one individual. They say Anthropic bypassed DRM protections and removed or altered copyright management information. Titles named include Don’t Be Cruel, Can’t Help Falling in Love and A Little Less Conversation. One oddity: the reporting refers to both 274 and 247 songs, so the exact count is unclear.
Lead attorney Richard Busch wrote that Anthropic has built its entire business model on infringement at a scale never seen before. That is advocacy, not fact. But the DRM and copyright-management claims matter more than the rhetoric. They go beyond the usual argument over whether training is fair use and target how the data was allegedly acquired. That is a harder thing to defend.
Anthropic is not alone in this. OpenAI and others face their own copyright suits. But Anthropic has had a rough run in music specifically. Sony and Warner recently sued over thousands of works. Before that, Universal Music and Concord Music Group sued over roughly 20,000 songs they say were illegally downloaded and used for training. Those cases seek damages in the billions.
So why does a $41 million claim matter? Because it is narrow. A single artist, one catalog, a clear per-song number. If it holds up, it gives other rights holders a template: pick a recognizable body of work, document the acquisition, and put a price on it. Estates and publishers with deep catalogs, think Beatles or Prince, could follow the same playbook.
Anthropic has argued that training on copyrighted music is fair use. It told the NYT it disagrees with the claims and will defend itself. Fair use has had mixed results in AI cases so far, and courts have treated how the material was obtained as a separate question from how it was used. That distinction is where Anthropic is most exposed here.
For developers and founders building on these models, the practical risk is indirect, not immediate. Lawsuits like this push up licensing costs, and licensing deals are already the alternative path. Competitors that signed music or content agreements early may look smarter in hindsight. For now, treat this as a real signal about data sourcing, not noise. The outcome will not change your API bill tomorrow, but it could shape what these models are allowed to learn from.
