Anthropic and major music publishers are making key arguments in a closely watched US copyright case that could help determine whether artificial intelligence companies can legally train models on copyrighted song lyrics without permission.
The case involves publishers including Universal Music Group, Concord and ABKCO, which accuse Anthropic of copying lyrics from songs associated with artists including Beyoncé, The Rolling Stones and The Beach Boys to train its Claude AI models.
US District Judge Eumi Lee in San Jose, California, is hearing arguments on whether Anthropic’s use of the material qualifies as fair use under US copyright law.
The outcome could influence dozens of similar lawsuits against AI companies.
FCRF’s Chief AI Officer Certification Returns on November 14 to Prepare Future AI Leaders
Publishers Say Claude Used Lyrics Without Licences
The music publishers argue that Anthropic copied copyrighted lyrics without obtaining licences and then used those works to build commercial AI systems.
Their complaint says Anthropic benefited from protected material while avoiding licensing costs normally paid by services that use song lyrics.
The publishers also argue that Claude can reproduce or generate material similar to copyrighted lyrics.
That creates a second concern.
Their case is not only about whether Anthropic copied material during training, but whether the resulting system can compete with markets where song lyrics are normally licensed.
The publishers say that allowing AI companies to use lyrics without payment could undermine legitimate licensing businesses and reduce incentives for songwriters and rights holders.
They are asking the court to reject Anthropic’s argument that training is protected by fair use.
Anthropic Says AI Training Is Transformative
Anthropic argues that training an AI model is fundamentally different from simply republishing a song lyric.
The company says the material is used as part of a much larger dataset to teach the model patterns in language.
Under that argument, Claude is not being created as a replacement database for individual songs.
Instead, the copyrighted material contributes to a system capable of performing many different tasks.
That is the core of Anthropic’s fair-use defence.
Fair use is a part of US copyright law that can allow copyrighted material to be used without permission in certain circumstances.
Courts usually examine factors including the purpose of the use, how much of the original work was copied and whether the new use harms the market for the original.
Anthropic says AI training is highly transformative because the model does not simply store works for conventional consumption.
The publishers strongly disagree.
US Courts Are Already Split on AI Training
The case arrives at a particularly important moment.
US judges have begun issuing different views on whether AI training qualifies as fair use.
In one earlier case involving Anthropic and authors, US District Judge William Alsup said training AI systems could be considered highly transformative.
But he separately found problems with Anthropic’s storage of pirated books in a central library that was not directly justified by model training.
That dispute later ended in a $1.5 billion settlement.
Another judge, Vince Chhabria, raised broader concerns in a separate case involving Meta.
He warned that generative AI could eventually flood creative markets with competing content and reduce the economic value of human work.
Those differing approaches have left AI companies and copyright owners without one clear legal rule.
The music-publisher case could help sharpen that boundary.
Market Harm Could Become the Decisive Issue
One of the most important arguments is likely to be whether AI-generated output harms existing markets for copyrighted music.
Anthropic says its models do not replace the original works.
The publishers argue that a model trained on lyrics can generate competing material and reduce demand for licensed lyric services.
That distinction could become central.
Courts often give significant weight to whether an allegedly transformative use damages an existing or potential licensing market.
The publishers say there is already a functioning market where companies pay to use song lyrics.
If AI developers can bypass that market, they argue, the economic harm is direct.
Anthropic counters that recognising such a broad licensing right could make large-scale AI training practically impossible because developers would need permission for enormous volumes of material.
That is one reason this case has implications far beyond music.
Ruling Could Shape AI Training Across Industries
The dispute could affect books, journalism, photography, films and other copyrighted material used in model training.
OpenAI, Meta, Google and several AI startups are already defending similar lawsuits.
If courts broadly accept fair use, AI companies could continue training models on large datasets without negotiating licences for every individual work.
If courts move the other way, developers could face higher costs and pressure to strike licensing agreements before using copyrighted material.
Some companies have already started doing that.
Publishers, media companies and entertainment groups increasingly see AI training data as a commercial asset rather than material that should be freely scraped.
The Anthropic case therefore sits at the centre of a much larger question: whether copyright law will treat AI training as a new kind of transformative use or as a commercial activity that requires permission.
Judge Lee’s eventual ruling may not settle the issue nationwide, but it could become another major precedent in a rapidly developing area of law.
What this means for you:
The case could affect how much AI companies must pay for the books, music, journalism and other material used to train future models. A ruling against broad fair use could make licensed training data far more important.
Follow for daily updates on cybercrime, corporate fraud, DFIR, hacking, investigations, and digital forensics
