Anthropic faces lawsuit over using Elvis Presley music to train Claude

The legal challenge begins

A group of rights holders has filed a lawsuit against the San Francisco-based artificial intelligence company Anthropic. The legal action, filed on October 23 in a federal court in San Jose, centers on the use of Elvis Presley's musical catalog. The plaintiffs allege that the company used copyrighted works without permission to train its Claude AI models.

The lawsuit specifically identifies 274 musical compositions that were allegedly used in the training process. These include famous hits such as 'All Shook Up,' 'Return to Sender,' 'Can't Help Falling in Love,' and 'A Little Less Conversation.' The plaintiffs represent various interests, including Authentic Brands Group, Raleigh Music Publishing, Abilene Music, and several individual trusts and people who control the rights to these songs.

Anthropic has publicly stated that it disagrees with the allegations made in the complaint. A company spokesperson noted on October 26 that the firm intends to defend the case fully. This litigation joins a growing list of high-profile disputes regarding how AI companies obtain and utilize copyrighted material to build their technology.

How the data was collected

The plaintiffs argue that Anthropic gathered copyrighted lyrics from across the internet to build its training sets. They specifically point to large datasets such as Common Crawl, The Pile, and the Internet Archive as sources for this material. The complaint claims these collections included Presley's lyrics without any authorization from the owners.

The lawsuit further alleges that Anthropic maintained a permanent internal repository of this copyrighted material. The plaintiffs contend that this data was kept for use in future training cycles and other various purposes. They argue that the copying of these works occurred at multiple stages of the development process, including when the material was first acquired, cleaned, and stored.

Additionally, the complaint accuses the company of stripping away important copyright management information. This includes removing song titles, composer and songwriter credits, publisher information, and official copyright notices. The plaintiffs claim these actions were taken as the material was incorporated into successive versions of the Claude models.

AI outputs and derivatives

The lawsuit also focuses on the specific types of content that Claude generates when users interact with it. The plaintiffs claim that the AI has produced lyrics from the Presley catalog and created derivative works based on those original songs. They argue that the model is capable of mimicking the core elements of the music without permission.

The complaint highlights an interaction where Claude allegedly created versions of songs specifically for a 'modern audience.' According to the plaintiffs, the resulting outputs retained the original songs' structures, hooks, and verse, chorus, and bridge arrangements. The AI only changed some of the wording while keeping the underlying composition intact.

The rights holders also claim that Anthropic implemented filters to prevent the generation of copyrighted material following previous lawsuits. However, they argue that these safeguards are insufficient because they can be bypassed by changing the wording of a user's request. They seek to prohibit further infringement and want the company to destroy unauthorized copies of the works found in training datasets, model weights, and server infrastructure.

A pattern of litigation

This case is part of a broader trend of music publishers and copyright owners challenging the use of their works in generative AI. Other music companies, including Universal Music Group and Concord Music Group, have already sued Anthropic over allegations involving copyrighted lyrics. These disputes reflect a fundamental disagreement over the legality of scraping massive amounts of data to train commercial models.

Anthropic's history with copyright litigation extends beyond the music industry. The company has also faced legal challenges regarding the use of copyrighted books to train its systems. Last year, the company reached a $1.5 billion agreement with a group of authors who claimed the company used pirated books for its training purposes.

The timing of this new lawsuit is also notable as it arrived one day after Anthropic submitted documents for a proposed initial public offering. As the company moves toward a public listing, these legal battles over intellectual property and data sourcing remain a significant hurdle. The outcome of the Presley case could have lasting implications for the business models of many AI developers.