Elvis Presley Rights Holders Sue Anthropic Over 274 Songs

What happened

Authentic Brands Group, a licensing group that controls Elvis Presley's music and image, filed a federal lawsuit against Anthropic, the company behind the Claude AI models, on 23 October 2026. According to Rolling Stone, the case is captioned ABE EPE LLC v. Anthropic and says the company infringed 274 Presley songs. The list runs from hits such as Can't Help Falling in Love and Don't Be Cruel to lesser-known tracks.

The complaint makes two separate claims. The first is that Anthropic used the recordings and compositions to train Claude without a licence. The second is that Claude produces output that reproduces or closely reworks the songs. Rolling Stone quotes the filing as saying Claude has created infringing output from virtually every song in every catalogue available online. Billboard first reported the case, and according to the reporting, the plaintiffs want $150,000 for each track, totalling about $41 million. The filing was viewed by Pitchfork, according to Yahoo's coverage. Outside the United States, British outlets converted the total to roughly £31 million, as AOL reported.

Anthropic has said it disagrees with the claims and intends to defend the case fully. The allegations have not been tested in court, and none of the claims has been decided.

Why it matters for businesses that automate work

Most companies that automate routine work do not train their own large language models. They rent tools and feed them documents, templates and data. That means the questions in this case reach them in two places: what goes into the tool, and what comes out of it. An output that reproduces protected lyrics or passages is a problem when it lands in a customer email, a marketing draft or a monthly report.

The damages figure is also worth reading carefully. The request is built on a per-work count. The table below shows how the reported numbers fit together.

The $150,000 figure is the ceiling that US law allows for wilful infringement. A court may award up to that amount per work, but it is not a sum already owed. Many cases end well below the ceiling. For a business, the lesson is that statutory damages scale with the number of works involved, so a workflow that repeatedly copies third-party material can carry more exposure than a single mistake.

The case also sits alongside a larger precedent. In 2025, Anthropic agreed to pay $1.5 billion to settle a separate class action brought by authors, which works out to about $3,000 for each of roughly 500,000 works, according to a Fieldfisher analysis published that year. That settlement concerned books, not songs, but it shows that AI training disputes can end in large payments.

What changes for a business that automates work, in practice

The lawsuit does not change what an AI tool is allowed to do today. It does change the checklist a careful owner or operations head should run before renewing or expanding an AI tool. Four steps are worth taking this quarter:

1. Ask each AI vendor in writing how it handles copyright claims about its outputs, and whether its business terms include any protection for customers.
2. Keep song lyrics, book extracts, licensed images and other third-party material out of prompts and templates unless your company holds the rights.
3. Record which workflow, tool and prompt produced each document that leaves the company, so a disputed output can be traced quickly.
4. Name a person who reviews outputs that go to customers, regulators or the public before they are sent.

Automation can run the repetitive steps and keep those records. It cannot decide whether a specific output infringes a specific work. A person still has to judge that, and a person should own the decision about which tools are approved.

This explains the rule in general terms; it is not legal or tax advice for your case.

What to watch next

Three things will show how far this case reaches. The first is Anthropic's formal response in court, which will show which defences it raises. The second is whether the court treats the training claim and the output claim differently, since they raise different questions. The third is whether other music publishers, estates or rights holders file similar complaints against AI developers. Until a court rules, the figures in this article are what the plaintiffs have asked for, not what anyone has been ordered to pay.