Anthropic: The Copyright Ruling Split on Training vs Piracy
As AI technology continues to change and develop, copyright cases involving AI are becoming increasingly common while governments struggle to establish effective regulations and enterprises grapple to stick to them.
In the last few months, the AI industry has faced a high amount of legal challenges over its use of copyrighted content.
Disney and Universal filed a lawsuit against AI image generator Midjourney, accusing the company of piracy, Reddit did the same with Anthropic – and the BBC is considering legal action over unauthorised use of its content by AI developers.
Now, a federal judge has delivered a split decision in a closely watched copyright case, with a ruling that allows Anthropic to continue using copyrighted books for training purposes whilst ordering the company to face trial over its use of pirated materials.
Anthropic’s trial over pirated book collection
Judge William Alsup of the US District Court rules that Anthropic’s training of its Claude large language model (LLM) using copyrighted works constitutes “exceedingly transformative” use under American copyright law.
The case emerged from a lawsuit filed by three authors against Anthropic, as mystery thriller writer Andrea Bartz, whose novels include: We Were Never Here and The Last Ferry Out, joined non-fiction authors Charles Graeber and Kirk Wallace Johnson in accusing the company of copyright infringement.
The writers claimed Anthropic used their work without permission to train Claude and build what they described as a multi-billion dollar business.
The judge’s decision is one of the first significant legal precedent addressing whether AI training constitutes fair use under copyright law.
However, Judge Alsup rejected Anthropic’s motion to dismiss the entire case, ruling that the company must stand trial over its maintenance of what he described as a “central library” containing more than seven million pirated books.
“Like any reader aspiring to be a writer, Anthropic’s LLMs trained upon works, not to race ahead and replicate or supplant them — but to turn a hard corner and create something different,” he writes in his ruling.
“If this training process reasonably required making copies within the LLM or otherwise, those copies were engaged in a transformative use.”
He further notes that the authors had not claimed the training process produced “infringing knockoffs” that replicated their original works for users of the Claude system.
He indicates such claims would constitute “a different case” with potentially different legal outcomes.
How Anthropic’s trial represents the bigger picture
Some AI companies have responded to the legal pressure by negotiating licensing agreements with content creators and publishers.
These deals provide explicit permission to use copyrighted material for training purposes, though they represent additional costs for AI developers.
The split nature of Judge Alsup’s ruling shows the complex legal questions surrounding AI development.
Whilst he endorsed the fair use defence for training activities, he distinguished between the transformative use of content for learning and the unauthorised copying of pirated materials.
As a result, Anthropic could face damages of up to US$150,000 per copyrighted work if found liable for using pirated copies.
The company maintains a library of millions of books according to court documents, raising potential financial exposure.
“We are pleased by the judge’s recognition that our use of the works was transformative, but we disagree with the decision to hold a trial about how some of the books were obtained and used,” Anthropic states.
“We remain confident in our case and are evaluating our options.”
Explore the latest edition of AI Magazine and be part of the conversation at our global conference series, Tech & AI LIVE. Discover all our upcoming events and secure your tickets today.
Also sign up to our free weekly newsletter for the latest insights and stories straight into your inbox.
AI Magazine is a BizClik brand