Elvis Presley's estate sues Anthropic for $41 million over AI training

The lawsuit against Anthropic

The owners of Elvis Presley’s music catalog have filed a federal copyright infringement lawsuit against the AI company Anthropic. The legal action was initiated in California on Friday, October 23, by Authentic Brands Group, the entity that owns the rights to the music and likeness of the artist. The plaintiffs are seeking a total of $48.6 million in damages for the unauthorized use of the work.

The complaint specifically names 274 songs from the Presley catalog as being used to train Anthropic's Claude models. Among the hits listed in the filing are well-known tracks such as "Can’t Help Falling in Love," "A Little Less Conversation," and "Don’t Be Cruel." The plaintiffs, which also include Raleigh Music Publishing and several trusts, are asking for the maximum statutory figure of $150,000 per infringed song.

Anthropic has officially rejected these allegations and stated that it disagrees with the claims made by the estate. The company announced that it intends to defend the case fully in court. This legal battle highlights the growing tension between AI developers and the entities that own massive amounts of historical cultural content.

Allegations of unlicensed output

The plaintiffs argue that Claude has produced infringing content from virtually every music catalog available on the internet. They claim the AI is capable of recreating Presley's music with high fidelity. The lawsuit suggests that the model is not just learning styles but is actually reproducing protected works in a way that violates copyright law.

A significant portion of the complaint focuses on the effectiveness of Anthropic's safety measures. The plaintiffs allege that even after the company implemented guardrails to prevent the generation of copyrighted material, a recent version of Claude still produced songs. When asked to recreate music for a modern audience, the AI reportedly produced outputs that the plaintiffs describe as unlicensed derivative works.

According to the complaint, the AI's results were described by the model itself as "landing very close to the originals." The plaintiffs argue that these outputs reused the same structures and hooks as the original songs with only minor changes to the wording. They point out that the AI included full sets of verses, choruses, and bridges, which they believe constitutes a direct infringement of the original compositions.

A shift in legal strategy

This case is notable because it represents one of the first high-profile copyright lawsuits brought by the owner of a single artist's catalog. In previous years, most major AI copyright litigation has come from large music publishers like Universal Music Group, Concord, and ABKCO. By filing this suit, the Presley estate is signaling a new era of litigation where individual legacies may take the lead in protecting their intellectual property.

The legal landscape for AI and music is becoming increasingly complex as different types of entities seek to establish their rights. While large publishers have focused on broad licensing and the use of lyrics, this suit focuses on the specific reproduction of a legendary artist's body of work. It highlights how the value of a single artist's catalog can be a primary target for legal action against generative models.

The lawsuit follows a separate copyright dispute involving Anthropic and a group of authors. In that instance, a judge ruled that training AI on lawfully acquired books constituted fair use, but using pirated copies did not. This distinction between legal and illegal acquisition of data remains a central theme in the ongoing debate over how AI models are built and what data they are allowed to consume.

The question of fair use

Whether the legal reasoning used in the book-related cases applies to copyrighted music remains an open question for the courts. The core of the debate rests on whether the training process for large language models qualifies as a transformative use of the data. If the court finds that music training is different from book training, it could significantly impact how AI companies source their training data.

The plaintiffs argue that the AI's ability to recreate the structure and hooks of the songs means it is creating unlicensed derivative works. They believe that the model is not merely learning a style but is essentially copying the core components of the music. This distinction is critical for determining whether the AI's output is a new creation or a copyright violation.

As the case moves forward, the outcome will likely influence how other artists and estates approach their rights in the age of generative AI. It puts Anthropic in a position where it must defend its training methods against specific, high-profile claims of infringement. The result could set a standard for the entire music industry as it navigates the intersection of technology and intellectual property.