Elvis Presley rights holders sue Anthropic over AI training on 274 songs

The lawsuit against Anthropic

The rights holders for Elvis Presley have filed a federal lawsuit against Anthropic in San Francisco. They allege that the AI company used 274 of Presley's songs to train its Claude models without obtaining any licenses or making payments. The claim covers a wide range of hits, from "All Shook Up" to "Can’t Help Falling in Love."

The plaintiffs, which include Authentic Brands Group and Raleigh Music Publishing, are seeking $150,000 in statutory damages for each song. This per-song figure is the maximum allowed under US copyright law for willful infringement. When multiplied across the 274 tracks, the requested amount exceeds $41 million.

The lawsuit also seeks roughly $50 million in total once court costs and attorney fees are added. The rights holders argue that Anthropic built its entire business model on infringing copyrights at a massive scale. They claim the infringement continues today whenever a user generates content that mimics Presley's lyrics or creates unauthorized adaptations.

Allegations of data theft

The complaint goes beyond the mere scraping of lyrics from the internet. The plaintiffs argue that Anthropic actively circumvented protective measures on the music files. They further allege that the company removed or altered copyright management information from the songs it used during the development of Claude.

Attorney Richard Busch stated that the plaintiffs act as guardians of the singer's work, which they describe as some of the most iconic ever made. This protective stance is backed by the fact that the catalog is managed as pure intellectual property. Because Presley died in 1977, there are no living artists involved in the licensing decisions, making the legal claim straightforward.

The lawsuit contends that the AI firm was informed that it was accessing protected content without authorization. Despite this warning, the plaintiffs claim the company continued its practices. This focus on "willfulness" is important because it may influence how a jury or judge decides to award the final damages.

A pattern of AI litigation

This legal battle arrives while Anthropic is already dealing with significant copyright repercussions. The company recently agreed to pay a record $1.5 billion to settle a class-action lawsuit brought by authors. Those authors claimed their books were pirated to train the Claude models, making it the largest publicly reported copyright recovery in history.

The situation is complicated by a split ruling from US District Judge William Alsup. The judge previously ruled that using books to train AI might not violate copyright law, but he ordered Anthropic to stand trial over the use of pirated copies. This highlights a messy legal landscape where the method of acquiring data matters as much as the training itself.

The Presley case is not isolated, as at least half a dozen similar suits regarding music in AI training are currently pending in the same San Francisco court. These cases collectively represent a push by the music industry to stop AI companies from treating the internet as a free buffet. Publishers view song catalogs as their primary business model and are fighting to ensure they receive compensation.

The path forward for AI

Anthropic has issued a brief statement saying it does not agree with the accusations and plans to fight the case in court. This is a firmer stance than the one taken in the authors' dispute, which the company chose to settle rather than litigate to a final verdict. The outcome of the Presley case will depend on several factors, including whether the plaintiffs seek an injunction to block future use of the songs.

Discovery in the case may reveal exactly how Presley's lyrics and compositions entered the training data. It will also show if the source of the data resembles the pirated-books problem that led to the $1.5 billion settlement. The court will have to decide where the line sits between fair use and copyright theft in the context of music.

Music adds a unique layer to this problem because lyrics and compositions are often licensed separately from the actual recordings. Rights are frequently fragmented across various publishers, trusts, and heirs. As the case progresses, it will likely shape how AI companies navigate the complex web of music ownership and licensing requirements.