Anthropic sued over alleged use of 274 Elvis songs in AI training

By Rebecca Stone

Authentic Brands Group is seeking about $41 million from Anthropic in a copyright lawsuit alleging the AI company used 274 Elvis Presley songs without permission to train Claude.

The suit, filed Friday and first reported by Billboard, identifies Authentic Brands Group as the licensing organization that owns Presley’s music and likeness. It ranks among the first major legal challenges brought directly by an artist’s estate against an AI developer.

What does Anthropic face?

The complaint says Anthropic used 274 Presley recordings as training material without authorization. The works listed include “Can’t Help Falling in Love,” “All Shook Up,” and “A Little Less Conversation.”

It also alleges that Claude has generated infringing musical output on a vast scale. The lawsuit states that “from its inception, Anthropic’s AI model, ‘Claude,’ has created infringing output of copyrighted material of virtually every song from every catalog available on the internet.”

That allegation extends beyond the 274 songs identified as training material. The plaintiffs are connecting Anthropic’s use of Presley’s catalog to Claude’s broader production of allegedly copyrighted material, according to the filing.

The quotation is broader than the three named songs. The complaint pairs a defined training set with an expansive claim about what the trained model can reproduce.

Richard Busch, the lead attorney, framed the case as an attack on Anthropic’s business model. He wrote: “The Elvis plaintiffs bring this action to enforce these exclusive rights against the defendant that has built its entire business model based on copyright infringement at a scale never seen before.”

What damages are at stake?

Authentic Brands Group is seeking $150,000 for each allegedly infringed track. Applied to 274 songs, the demand produces the reported total of about $41 million, or about £31 million.

The complaint also focuses on what Anthropic knew about its safeguards. It says the company instituted guardrails to limit outputs that infringed music copyrights, but that those measures failed to prevent the conduct described in the lawsuit.

The filing says Anthropic continued generating infringing material after being caught and knew its guardrails were ineffective. It puts that knowledge at the center of its allegations:

“Even Anthropic’s institution of guardrails (albeit ineffectively) to purportedly limit outputs which infringe upon music copyrights, after being caught, along with its knowledge that its guardrails were ineffective to stop creating outputs which infringed music copyrights, demonstrates that Anthropic knowingly engaged and continues to willfully engage in copyright infringement of others’ musical works,” the lawsuit states.

The complaint therefore seeks to characterize Anthropic’s conduct as willful infringement, not merely unauthorized use. It says Anthropic knew its safeguards did not stop the outputs at issue.

By joining the 274 songs, Claude’s output and Anthropic’s alleged knowledge, the filing supports its contention that the conduct was knowing and willful.

Why does the case matter for AI music?

Copyright claims against AI companies already include the Recording Industry Association of America and the labels Sony, Universal and Warner Records.

The Presley case stands apart because it comes directly from the organization that owns one artist’s music and likeness. The report describes it as one of the first major legal challenges brought directly by an artist’s estate against an AI developer.

Its outcome could set a broader precedent for how AI companies may use existing music catalogs. The dispute therefore reaches beyond Presley recordings. It places the legality of copyrighted music in AI training and output before the courts through a high-value claim tied to one artist’s estate.

The damages demand also gives other rightsholders a concrete figure to watch: $150,000 for every track the plaintiffs say Anthropic infringed. The $41 million total depends on the full 274-song count in the complaint.

That structure combines an identifiable catalog with a per-work monetary demand. The filing does not leave the alleged infringement at the level of models or music in general; it names recordings and assigns a monetary amount to each one.

How do AI music platforms use recordings?

The report describes AI music platforms as systems that analyze large amounts of recorded music to learn sounds, structures and expressions. That analysis lets a model reproduce or recombine musical patterns from a training set.

Those platforms then let users create AI-generated music through text or speech prompts. The recorded catalog supplies the source material; the prompt directs the model’s output.

That workflow puts rights clearance for training music at the center of the business and legal dispute. The Presley complaint alleges that Anthropic used protected songs without permission and that Claude produced infringing output from protected works.

The complaint does not limit its argument to the three named examples. Its broader output allegation and its 274-track training allegation cover a larger body of claimed infringement.

In the plaintiffs’ framing, the catalog serves as both input and evidence. The songs show what Anthropic allegedly used to train Claude, while the output allegation supplies the claimed harm beyond Presley’s three listed titles.

What are musicians concerned about?

The report records repeated concerns among musicians about two outcomes: their music and voices being used without permission, and AI-generated “slop” replacing human-made songs. The Presley lawsuit turns the first concern into a specific claim involving a named artist, a finite catalog and a seven-figure demand per work.

The report also notes that some AI bots have begun impersonating real living artists on music streaming platforms. That puts a separate identity problem alongside the copyright claims: an AI system can present music under an artist’s name, not merely copy material used to train it.

Spotify announced the scale of one cleanup last week. The company said it had removed about 50 million songs considered AI slop from user recommendations over the previous year.

Spotify’s 50 million removals vastly outnumber the 274 recordings targeted by the Presley lawsuit. The figures measure different problems, but both show AI-generated music reaching major streaming interfaces at scale.

Spotify’s announcement came last week; the Presley lawsuit was filed Friday. The report therefore places a large-scale platform cleanup and a targeted legal claim within the same recent period.

Could the ruling reshape training rules?

The case concentrates the plaintiffs’ theory into a defined set: 274 identified songs, claimed use without consent, alleged Claude output and alleged knowledge that safeguards failed. A court would confront the complaint’s account of both training and model behavior.

The filing does not settle whether the catalog use was authorized or whether Claude’s safeguards were sufficient. Those contested points sit at the center of the suit. Its result could set a broader precedent for how AI companies are allowed to use existing music catalogs.

For AI developers, Friday’s filing turns permission for recorded catalogs into a potential $150,000-per-track legal exposure, and an adverse ruling could make the case a reference point for model training on music and liability when safeguards fail.