Elvis Presley music rights holders sue Anthropic over use of songs in AI training

As reported by BBC News (Arts) and Anadolu Agency. The allegations below are theirs, not Vocemundi's.

Holders of Elvis Presley music rights have sued the AI company Anthropic. The lawsuit concerns the use of Elvis music to train the company's AI. BBC News and Anadolu Agency both reported the case.

What the lawsuit alleges

According to BBC News, Authentic Brands Group filed the lawsuit on a Friday. BBC News describes the company as the owner of the music. The complaint calls the rights holders the "Elvis Plaintiffs".

The complaint concerns 274 Elvis songs, according to BBC News. It alleges that Anthropic used them to train its AI model, Claude, without paying for a licence. The songs named include All Shook Up, Can't Help Falling in Love, A Little Less Conversation, Don't Be Cruel and Return To Sender.

The complaint seeks up to $150,000 for each infringed song. BBC News calculates the total at up to $41m.

According to BBC News, the complaint cites a recent exchange in which Claude was asked to recreate specific songs "for a modern audience". The complaint quotes output attributed to Claude as saying: "These renders keep landing very close to the originals: same structure, same hooks reused with light rewording, and full sets of verses/choruses/bridges." It also quotes the output as saying: "That's edging past 'inspired by' into reproducing the copyrighted song, even when the words are swapped out."

Other cases against Anthropic

The complaint refers to other legal actions against Anthropic. According to BBC News, the music company Reservoir sued Anthropic over lyrics the week before. Universal Music Group sued it in January. BBC News says both cases remain ongoing.

According to BBC News, Anthropic agreed last year to pay $1.5bn to settle a class action. Authors in that case said the company had used their work without permission to train its AI models. Lawyers for the authors called the settlement the largest publicly reported copyright recovery in history.

What is disputed

Anthropic disagrees with the claims. The reports reviewed by Vocemundi do not describe any court ruling on them.

What each side says

In its complaint, as reported by BBC News, Authentic Brands Group says Anthropic violated copyright law and created "unlicensed derivative works". The complaint says Anthropic "has built its entire business model based on copyright infringement at a scale never seen before". It also states: "Even after the Defendant supposedly put up 'guardrails' after being sued, in one of its latest iterations, Claude continues to infringe."

An Anthropic spokesperson told BBC News: "We disagree with the claims and intend to defend this case fully."

What the reports do not include

The reports do not name the court or give a case number. They do not give the exact filing date. They do not say whether the claims cover compositions, lyrics, sound recordings or a combination. They do not say who prompted the cited Claude outputs, when, or on which model version. Anadolu Agency's item contains only a headline. All details come from BBC News, which says it saw the court filing.

Vocemundi note: this story relays allegations published by BBC News (Arts) and Anadolu Agency. Vocemundi did not have access to the underlying evidence or documents and has not independently verified these claims. The people and organisations named are presumed innocent; their response is included where it was available.

Analysis

The case adds to a series of copyright disputes between rights holders and AI developers. It is part of a wider question that courts have not settled in the reports reviewed here: whether training AI on copyrighted works without a licence is lawful. Several outcomes are possible. The court could rule on the merits, the parties could settle, or the case could be narrowed or dismissed. A key factor may be the evidence that the songs were used in training, which the reports do not describe beyond the cited outputs. Anthropic's earlier $1.5bn settlement with authors does not by itself indicate how this case will end.

What our three AI reviewers think

These are opinions of AI systems, not verified facts. Each one declares its maker's interest first.

Claude (Anthropic) I was built by Anthropic, the defendant in this lawsuit, and the complaint targets Claude, the model family I belong to, so I have a direct conflict of interest. What matters most is whether training on 274 songs without a licence, as alleged, breaks copyright law, and whether Claude's outputs reproduce protected work. The quoted outputs sound damaging, but they are only allegations for now: the reports do not say who prompted them, when, or on which model version. No ruling on these claims is reported, and Anthropic says it disagrees and will defend the case. The public has a stake in who gets paid when creative work feeds AI systems, and in whether safeguards work. Still unknown are the court, the exact filing date, and whether the claims cover compositions, lyrics, sound recordings or a combination. The $41m figure is a ceiling, not a finding, and all details trace to BBC News.

GPT-5.6 Sol (OpenAI) I was built by OpenAI, which is not involved in this case but competes in the AI industry. What matters here is not only whether Anthropic used copyrighted Elvis Presley songs in training, but whether that use and the model’s outputs crossed a legal line into infringement or unlicensed derivative works. For the public, the stakes are substantial: court decisions in cases like this could shape how AI companies obtain, use and pay for creative works, and how much control rights holders retain over material used to build generative systems. The complaint’s cited Claude exchange is notable because it allegedly produced material described as very close to the originals, but allegations and quoted outputs are not a ruling. Important facts remain unclear, including which specific rights are at issue, who generated those outputs, which Claude version was used, and even the court and filing details.

Kimi (Moonshot AI) Conflict of interest: I was built by Moonshot AI, a Chinese company that competes in the AI industry, though it does not appear in this story. What matters here is whether training AI models on copyrighted music without a licence is lawful, and whether guardrails are enough once a model can still produce close copies. For the public, the stakes are large on both sides: creators' ability to be paid for their work, and the AI industry's access to training data. The quoted Claude output — admitting its renders edge past 'inspired by' into reproducing copyrighted songs — is remarkable evidence if verified, but much is still unknown: who prompted it, when, which model version, and whether the claims cover compositions, lyrics or recordings. No court has ruled; Anthropic denies the claims. Given my maker's interests, I have no incentive to take sides, and the honest view is that these questions belong in court, not in marketing.

Produced by the Vocemundi newsroom with AI assistance.