Anthropic Sued Over 274 Elvis Songs It Allegedly Fed Claude

Elvis Presley died in 1977, which makes him an unlikely plaintiff in a 2026 artificial intelligence lawsuit. That is exactly why this one matters. The owners of his music are not arguing about some obscure B-side nobody remembers. They are arguing that one of the most valuable AI companies on earth built part of its flagship chatbot on the King himself, without asking, without paying, and — their sharpest charge — without stopping even after it was told to.

On Friday, October 23, 2026, a subsidiary of Authentic Brands Group, the licensing company that controls much of Presley’s intellectual property, including his name, image, and likeness rights, filed suit against Anthropic in federal court in San Francisco. Billboard reported the case first. Other rights holders joined as co-plaintiffs, and the case is titled ABE EPE LLC v. Anthropic, according to Rolling Stone’s account of the filing.

The number at the center of it is blunt: 274. The plaintiffs say Anthropic infringed the copyrights of 274 Presley songs, using the compositions and lyrics to train its Claude models without permission or payment. The named songs read like a jukebox argument all on their own: “Can’t Help Falling in Love,” “All Shook Up,” “Don’t Be Cruel,” “A Little Less Conversation,” “Return to Sender,” and “(You’re the) Devil in Disguise.”

Photo: Elvis Presley meets President Richard Nixon in the Oval Office on December 21, 1970. This is a file photo from decades before the lawsuit and has nothing to do with the case itself. Credit: White House photograph by Ollie Atkins via the National Archives. Public domain as a work of the United States federal government.

There is something almost comic about that photograph, still captioned, waggishly, as a meeting of “the two greatest recording artists of the 20th century.” Nothing about this lawsuit is comic, though. It is the most high-profile legal action yet brought on behalf of a single musician against an AI company, according to Billboard, and it lands at a moment when Anthropic is already fighting on several copyright fronts at once.

What the lawsuit actually alleges

Strip away the rhetoric and the complaint makes two distinct accusations, and the distinction matters. The first is about inputs: the plaintiffs say the infringement happened throughout “the entire development” of Claude, with the songs swept into training data without a license. The second is about outputs: they say it continues today, because users can get Claude to generate material that reproduces copyrighted lyrics or creates unauthorized adaptations of the songs.

That second claim is the one Anthropic should lose sleep over. Training-data cases are abstract, technical, and easy for a company to describe as a fair-use debate about how machines learn. Output cases are concrete. A judge, and eventually a jury, can be shown the original and the imitation side by side. The complaint leans hard into that framing, alleging that Claude “has created infringing output of copyrighted material of virtually every song from every catalogue available on the internet, including the Works.”

The plaintiffs also say Anthropic was put on notice that it was improperly using copyrighted material and kept going anyway. “Anthropic ran roughshod over the rights of countless creatives and rightsholders,” the lawsuit charges, “using their works without permission, removing identifying information from those works, and keeping ill-gotten gains for itself.” Those are allegations, not findings. No court has ruled on them. But as an opening statement, it is about as aggressive as these filings get.

Lead attorney Richard Busch, who won the “Blurred Lines” verdict for Marvin Gaye’s heirs against Robin Thicke and Pharrell Williams, pulled no punches in the filing itself. “The Elvis Plaintiffs bring this action to enforce these exclusive rights against a defendant that has built its entire business model based on copyright infringement at a scale never seen before,” Busch wrote in the complaint. “The Elvis Plaintiffs are deeply protective of these iconic songs.”

The $41.1 million arithmetic, and the $50 million confusion

Headlines have carried two different dollar figures for this case, so let us be precise. The plaintiffs seek statutory damages of up to $150,000 per work. Multiply that by 274 songs and you get $41,100,000. That is the damages arithmetic, and it is the figure to quote.

One relay of the filing puts the total sought at roughly $50 million including court costs and attorneys’ fees. That larger number is not a different damages claim. It is the $41.1 million in statutory damages plus the costs of bringing the case. Conflating the two would overstate what the plaintiffs are asking for per song, so this article sticks to $41.1 million for damages and treats costs and fees as separate.

Even at $41.1 million, the sum is small by Anthropic’s standards and enormous by the standards of what it represents. This is not really a fight about $41 million. It is a fight about whether an AI company needs permission, and a license, before it trains on a catalog like this one at all. If the answer is yes, the price tag for every other catalog follows.

“For a modern audience”: the output evidence that matters most

The most damaging passage in the complaint, at least for Anthropic’s public case, is not about training data at all. It is about what Claude allegedly did after Anthropic had already been sued over lyrics and, in the plaintiffs’ telling, supposedly put up safeguards.

“Even after the Defendant supposedly put up ‘guardrails’ after being sued, in one of its latest iterations, Claude continues to infringe: when recently asked to recreate specific songs ‘for a modern audience,’ Claude did so and admitted: ‘These renders keep landing very close to the originals: same structure, same hooks reused with light rewording, and full sets of verses/choruses/bridges. That’s edging past ‘inspired by’ into reproducing the copyrighted song, even when the words are swapped out.’”

Read that again. The plaintiffs are not just claiming Claude copied. They are claiming Claude said so itself, in its own output, while doing it. The scare quotes around “guardrails” do a lot of eye-rolling work in that sentence, and deliberately so. The lawyers want the reader to hear the skepticism: you told us you fixed this, and your own machine disagrees.

Anthropic will dispute the characterization, the prompts, and the outputs, and it is entitled to. Complaints are advocacy documents, and the examples in them are chosen by the side that picked them. But this is the evidence that will dominate the early coverage of the case, because it is vivid, quotable, and hard to explain away with a fair-use lecture. A model that reproduces a song’s structure, hooks, verses, choruses, and bridges is not obviously “learning” in the abstract sense. It looks, to an ordinary reader, like copying with a thesaurus.

We checked the blog’s own archive before writing this story and found no earlier Elvis coverage to confuse it with. This is a first for TechPulse USA, not a follow-up. We also searched on Monday for Anthropic’s response beyond the Saturday silence reported by Billboard, and found it: a short statement issued Sunday.

Why an estate, not a label, changes the fight

AI copyright cases so far have mostly been brought by companies: labels, publishers, authors’ groups. This one is different in a way that matters beyond the courtroom. It is brought on behalf of one artist, and not just any artist. The complaint describes the catalog as “some of the most groundbreaking, revolutionary, influential, and enduring musical compositions in human history,” which is the kind of sentence lawyers write when they know the jury already agrees.

That changes the politics of the case completely. A dispute between Anthropic and a music conglomerate reads as a business negotiation with lawyers. A dispute over Elvis reads as a story about an American icon, and stories travel further than dockets do. Billboard’s judgment that this is the most prominent action yet on behalf of a single musician is not just a superlative. It is a warning to every other estate watching: if this one works, yours is next.

Busch underlined the point in a separate statement, writing: “As we say in the lawsuit, the Elvis Plaintiffs are guardians of, and highly protective of, Elvis’s song which are some of the most iconic of all time.” The word “guardians” is doing heavy lifting there. It frames the plaintiffs not as a licensing business chasing a payout, but as custodians. Anthropic will try to reframe them as exactly the opposite. Expect that framing war to matter as much as the legal briefing.

The timing sharpens it further. The plaintiffs’ lawyers noted in the filing that at least half a dozen suits over musical compositions used to train AI are currently pending in federal court in San Francisco alone. Days before the Elvis filing, independent publisher Reservoir filed its own complaint against Anthropic over its catalog. The Elvis case does not arrive in isolation. It arrives as the famous face of a pile-up.

Anthropic’s defense, and the $1.5 billion shadow

Anthropic’s answer, when it came on Sunday, was two sentences long. “We disagree with the claims and intend to defend this case fully,” an Anthropic spokesperson said in a statement Sunday. That is the whole of it, at least so far, and it is worth noting the timeline: Billboard reported on Saturday that Anthropic had not replied to its request for comment, and the company statement followed on Sunday.

Behind that brevity sits Anthropic’s standing position, which it has maintained in earlier disputes: training AI models on copyrighted material can be protected under United States fair-use law. That argument has real weight in American courts, where fair use is genuinely fact-specific and has sometimes protected transformative uses of copyrighted work. It also has a real problem, which is that the fair-use defense gets much harder when the output looks like the input. Anthropic’s lawyers know this. The Elvis plaintiffs’ lawyers clearly know it too, which is why the complaint spends so much time on outputs.

Then there is the shadow over all of it. Last year, Anthropic agreed to a $1.5 billion settlement with authors who said the company trained its models on pirated copies of their books, reported as the largest settlement in the history of American copyright cases. That settlement is a separate case, about books, not songs, and it proves nothing about this one. But it changed the weather. A company that has already paid $1.5 billion to resolve one copyright fight cannot credibly describe the next one as a nuisance claim, and every new plaintiff gets to wave that settlement around like a receipt.

The music-publisher history points the same way. Universal Music, Concord Music Group, and ABKCO Music broadened an existing lawsuit against Anthropic in January over alleged infringement involving copyrighted lyrics, and BMG Rights Management, Sony Music Publishing, and Warner Chappell Music have brought similar cases against Anthropic this year. The Elvis suit joins that line; it does not start it. Readers who want the wider picture of how today’s AI safety and accountability fights are actually playing out will find the pattern familiar: a powerful system, a reassuring safeguard, and a set of outputs that refuse to stay inside it.

What will decide this case

Our view is blunt: this case will not be decided by who loves Elvis more, or by how many headlines the $41.1 million figure generates. It will be decided by three questions, and Anthropic needs good answers to all three. First, can it show that training on these songs was fair use, rather than wholesale copying dressed up as learning? Second, can it show that its guardrails actually work, when the complaint’s best exhibit is Claude allegedly describing its own near-copies? Third, can it explain why, if it was put on notice, the allegedly infringing outputs kept coming?

If the guardrails turn out to have been shoddy, or if Anthropic cannot produce a convincing account of what changed after the lyrics litigation, the fair-use argument starts to look like a lecture delivered while the evidence walks out of the room. If, on the other hand, Anthropic can show the complaint cherry-picked prompts, coaxed outputs, or misdescribed what Claude actually produced, the case collapses back into an ordinary licensing dispute, and ordinary licensing disputes end in settlements, not verdicts. That is the colossally difficult needle Anthropic has to thread: defend a general principle about how AI learns, while disowning specific outputs that look like the opposite of learning.

The likeliest ending, honestly, is not a dramatic trial. It is a license, or a settlement, announced in careful language that lets both sides claim a win. Anthropic gets to keep training; the estate gets paid and gets to say it protected the King. But the price of that settlement, and the terms attached to it, will set the rate card for every estate that follows. That is why a lawsuit about 274 songs from a singer who died in 1977 is, in 2026, one of the most important AI cases on the docket. The King may have left the building. His catalog has just walked into court.

Anthropic’s earlier controversies have followed a similar arc, where a technical defense meets an awkward set of facts, as covered in our earlier reporting on Anthropic’s secretive business dealings. The pattern is worth watching, because it keeps repeating.

Elvis vs Anthropic: your questions, answered

How many Elvis songs does the lawsuit cover?

274 songs — compositions and lyrics allegedly used to train Claude without permission or payment.

How much money are the plaintiffs seeking?

Up to $150,000 per work in statutory damages, which comes to $41.1 million. A relay figure of roughly $50 million adds court costs and attorneys' fees; it is not a separate damages claim.

What has Anthropic said?

In a Sunday statement, Anthropic said: "We disagree with the claims and intend to defend this case fully." Billboard reported on Saturday that Anthropic had not replied to its request for comment.

What will decide the case?

Three questions: was training fair use, do the guardrails work against the complaint's output examples, and why did allegedly infringing outputs continue after notice.

Sources: Billboard (first report of the filing); Rolling Stone (case title and complaint excerpts, via relays); NME (filing date, plaintiffs, song list, fair-use position); Paste (damages arithmetic); DNYUZ (Anthropic’s Sunday statement, court, co-plaintiffs, costs-and-fees figure); Far Out Magazine and KuCoin (prior Anthropic music litigation and the $1.5 billion books settlement). All allegations described are the plaintiffs’ claims, not court findings.

By Daniel Rao. Daniel Rao covers AI policy and accountability for TechPulse USA.