Anthropic Faces Music Publishers in AI Copyright Dispute
Anthropic is facing major music publishers in a US court over how it trained its Claude AI system. The publishers claim Anthropic used their copyrighted lyrics without permission. Anthropic argues that US “fair use” rules cover what it did and that point is at the heart of the case.
On October 21, 2026, US District Judge Eumi Lee heard arguments in San Jose on the competing fair-use and market-harm positions. The case also asks a bigger question: how should US copyright law apply when AI systems are built using protected creative works?
This article explains what the publishers allege, how Anthropic is defending its use of the lyrics, why market harm matters and what the case could mean for AI and copyright law.
The Dispute Overview
The dispute is about how Anthropic trained Claude. In 2023, Universal Music Group, Concord and ABKCO filed a lawsuit against Anthropic. They claim the company used their copyrighted lyrics to teach Claude how to respond to user prompts. The songs involved are linked to artists like Beyoncé, the Rolling Stones and the Beach Boys. These are allegations not proven findings.
Song lyrics are creative works that copyright protects, so this case is about more than whether AI can write text. The court has to decide if building an AI system with lyrics, without permission, is allowed and if fair use covers it. Either way, the ruling will not decide whether all AI training is legal or illegal.
What Happened at the October 21 Hearing in the Anthropic Case?
On October 21, 2026, Judge Eumi Lee held a hearing in San Jose. Both sides asked her to decide key issues without a full trial. The hearing examined how the parties competing positions applied to Anthropic’s use of copyrighted lyrics and Claude’s ability to generate lyrics. The court also looked at whether Claude’s ability to generate lyrics harms the publishers’ markets.
Judge Lee did not announce a decision at the hearing. She considered the arguments so nothing has been decided yet. The court is still reviewing the legal and factual arguments from both sides. Her ruling could shape how courts approach AI training under US copyright law.
Also Read: OpenAI sued over “Astra” Trademark
How Anthropic Is Defending Its Use of Copyrighted Lyrics?
Anthropic argues that using song lyrics to train Claude counts as fair use. Its main argument is that the use is “transformative,” meaning the lyrics are used for a new purpose rather than copied to replace the originals. Anthropic says training helps Claude understand human language and supports progress in science, business, and education. Whether that use is truly transformative is up to the court.
On market harm, Anthropic says the publishers have not shown a concrete example of harm from Claude. Anthropic also pointed to earlier UMG statements that AI can be good for the music business. Anthropic relies on these points in support of its fair-use position.
The Publisher’s Side: Their Main Arguments Against Anthropic
The publishers argue that Anthropic’s use goes further than teaching Claude how language works. In their view, Claude can produce lyrics that compete with songs written by people. They point to this alleged ability to claim that the AI system could hurt existing markets for licensed lyrics and related creative works.
The publishers also said Claude received nearly 120,000 requests for ‘synthetic’ lyrics in six months. They treat this as a sign that AI-written lyrics could replace licensed human work, and they use it to support their market-harm argument.
Their larger concern is economic: AI-generated content could compete with licensed songs. Still, a business concern is not the same as a legal one. The court still has to decide whether that concern meets the requirements of a copyright claim.
Why Fair Use Is the Key Issue in the Anthropic Case?
Fair use is a US copyright rule that lets people use protected works without permission in certain cases. It does not automatically cover AI training. Courts review the facts and the law in each case before they decide.
Two questions stand out here. The first is purpose. Anthropic says training gives the lyrics a new technical purpose, which it describes as transformative. The publishers disagree and say this use is neither transformative nor fair.
The second issue is market harm. The court’s questions about Claude’s ability to write lyrics matter because, if an AI tool can create competing material, its impact on the market becomes a key fact. For now, only the publishers allege that harm. The court has not found it.
The Role of Earlier AI Copyright Rulings in the Anthropic Case
Other federal judges in Northern California have already looked at AI training and fair use and they did not all take the same path. Retired US District Judge William Alsup ruled for Anthropic in a case over training AI on books but his decision drew a line between pirated copies and ones bought lawfully. Two days later, US District Judge Vince Chhabria ruled for Meta in a similar case. He also warned that AI training may not be fair use in many situations.
These rulings show how unsettled the law still is. Neither one controls Judge Lee’s decision. Each case has its own facts and its own copyrighted works, and song lyrics are not books. That makes it hard to predict how this case will end.
What the Anthropic Decision Could Mean for AI and Copyright Law?
Judge Lee’s decision could reach well beyond the music industry. Authors, news organizations, visual artists, music rights holders and AI developers are all involved in lawsuits over AI training. The case could be important for many of those disputes. A fair use ruling might influence how other courts handle similar claims, even though it would not directly bind them.
For copyright owners, the case shows why it pays to watch how AI companies get and use protected works. The result may help shape the legal rules on using copyrighted material without permission to train AI.
AI developers face their own questions. They may need to think carefully about copyright risks in their training data and in what their models produce. Right now, no specific compliance rule is required unless a court or a law sets one.
Why the Anthropic Case Is Not Settled Yet?
The October 21 hearing did not end with a final ruling. Judge Lee has not decided whether Anthropic’s use of the lyrics is infringement and she has not decided that it is fair use either. Those questions are still undecided.
The case therefore should not be read as establishing a universal rule for AI training; its legal significance will depend on the specific facts, copyrighted works, evidence and reasoning considered by the court. Any wider effect should be seen as possible significance, not settled law. Until the court rules, the outcome remains uncertain.
The Broader AI Copyright Landscape
Across the US, courts are applying existing copyright principles to disputes involving the use of protected works in AI development. The cases differ in the types of works involved, the way the material was obtained, used and the evidence presented by the parties. Those differences make it difficult to draw a single rule from the decisions so far.
US courts have not settled on one approach. Judges in earlier cases looked at fair use differently and each ruling depended on its own facts. That leaves developers and rights holders without a clear answer. The music publishers case adds another view this time involving song lyrics but it is one part of a larger legal picture.
The Anthropic case comes down to one clear disagreement. The dispute therefore places two competing interests before the court: the use of copyrighted works in AI development and the protection of markets built around those works.
As of the October 21 hearing, Judge Lee had not ruled. Still, the case matters because it may add to the growing legal framework for copyrighted works and AI training.
From AnalystIP’s perspective, AI developers, copyright owners, publishers, creators and legal professionals will be watching closely to see how courts apply established copyright principles to AI systems.
