Elvis Presley rightsholders sue Anthropic over 274 songs allegedly used to train Claude

Elvis Presley never recorded a note for a chatbot, but his songbook is now at the centre of one of the sharpest copyright fights over AI training data. The companies and trusts that control rights in hundreds of songs he made famous have sued Anthropic, arguing that its Claude models were built on lyrics taken without permission or payment.

The suit was filed on Friday, October 23 in the US District Court for the Northern District of California, and it covers 274 musical compositions. It arrives as Anthropic heads for the public markets — and it asks a court to treat how AI companies store and reuse copyrighted work as a violation in its own right.

What the complaint alleges#

The lead plaintiff is ABG EPE IP LLC, a unit of Authentic Brands Group, which controls Presley’s name and likeness and holds a share of the publishing rights in songs he recorded. It is joined by Raleigh Music Publishing, Abilene Music, and trusts connected to the Aberbach family and to Luigi Creatore, who co-wrote “Can’t Help Falling in Love.”

The works named read like a greatest-hits list: “All Shook Up,” “Can’t Help Falling in Love,” “Don’t Be Cruel,” “A Little Less Conversation” and “Return to Sender,” among 274 compositions in total. The plaintiffs say Anthropic copied the lyrics and underlying compositions without a licence, kept them in a permanent, general-purpose library, and drew on that library again and again to train successive generations of Claude.

The complaint goes further than the training step. It alleges that Claude, when prompted, produces copies of the songs and new works derived from them, and that Anthropic stripped or altered songwriter credits and copyright information and sidestepped access controls. One distinction matters: the case concerns the compositions — the words and music, lyrics in particular — rather than Presley’s recordings, voice or likeness, which is why the plaintiffs are publishers and rights trusts rather than a record label.

The money, and the other demands#

The headline figure is $41.1 million: the maximum statutory damages of $150,000 per work, multiplied across 274 works, for what the plaintiffs characterise as wilful infringement. Added claims over removed copyright information ($6.85 million) and circumvented access controls ($685,000) take the total sought past $48.6 million before legal costs.

Money is only part of what is being asked for. The plaintiffs also want an injunction against further infringement and the impoundment or destruction of unauthorised copies. If a court granted that request in full, it would reach beyond damages into the training material itself.

Why this suit stands out#

The complaint identifies at least six pending musical-composition suits against Anthropic in the same district. What it adds is a clean, famous catalogue and a theory that attacks the whole pipeline at once: acquisition, storage, training and output.

The storage argument is the sharpest edge. The complaint leans on a July 2026 ruling in the book-authors litigation — the case Anthropic settled for $1.5 billion — in which a judge held that keeping unlicensed copies indefinitely in a general-purpose library is not protected as fair use. If that reasoning carries over to lyrics, then even a company that wins the argument about training itself could still lose on the library it trained from.

One caution: the filing’s concrete examples of chatbot output are drawn from other publishers’ cases, and its claim about Presley songs is made on information and belief. The storage and copying allegations stand on their own, but the “Claude sings Elvis” part of the story is, for now, an allegation to be tested rather than an established fact.

The wider music fight#

Universal Music Group, Concord and ABKCO sued Anthropic in October 2023 over 499 works, then expanded their claims in January 2026 to more than 20,000 works and more than $3 billion. BMG followed in March 2026, and Sony Music Publishing, Warner Chappell and Round Hill filed in August 2026. Anthropic has argued throughout that training on copyrighted material can qualify as fair use, and it pressed that position at a hearing in the publishers’ case on October 21 — two days before the Presley filing landed.

Publishers have an advantage book authors did not: short, distinctive works that are easy to match against model output, and statutory damages arithmetic that needs no proof of lost sales.

Anthropic’s response, and what to watch#

Anthropic responded on Sunday, October 25. A spokesperson said: “We disagree with the claims and intend to defend this case fully.” That frames the central collision: rights holders call AI training industrial-scale copying; AI companies call it transformative use.

Three things will decide how much this case matters. First, Anthropic’s answer and any motion to dismiss: whether it attacks the library theory head-on or tries to narrow the case to training alone. Second, whether the court treats the permanent library as a separate infringement, which is the ruling other rights holders are really waiting for. And third, the licensing ripple: every month these cases stay alive, the price of a music licence for AI training rises.

For now, nothing has been proven. The complaint is an allegation, Anthropic disputes it, and the songs are now Exhibit A in the argument over what AI is allowed to learn from.