Authors win $1.5bn payout after company used pirated works to train its AI model

Saturday, 6 September 2025

Claude, Anthropic’s rival to ChatGPT, was trained using pirated books.

In one of the largest copyright payouts in history, Anthropic AI has agreed to pay US$1.5 billion (NZ$2.54bn) to settle claims it built its chatbot on millions of pirated books.

If approved by a San Francisco court next week, the deal will compensate authors an average of US$3000 (NZ$5000) per book across roughly 500,000 works — a scale never before seen in the ongoing legal fights over how artificial intelligence uses creative content.

The case, brought in 2024 by authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson, accused Anthropic of copying millions of digitised books: including their own.

The company used these words to train Claude, its ChatGPT rival. Anthropic admitted it had relied on The Pile (a massive online dataset used to train AI language models), as well as scanning physical books.

Judge William Alsup’s June ruling delivered a split decision: training AI on legally acquired books was “exceedingly transformative”, and considered fair use.

However, using pirate sites to build datasets was not, Alsup ruled. It was this point that was headed for trial, with Anthropic facing potential damages of up to US$150,000 (NZ$254,000) per book, across 7 million works.

However, Anthropic struck a deal with the authors, marking what many see as a turning point.

Tech lawyer Cecilia Ziniti said the settlement paved the way for a licensing scheme for training data.

“It’s not the end of AI, but the start of a more mature, sustainable ecosystem where creators are compensated.”

Author groups hailed the payout. “AI companies cannot simply steal authors’ creative work just because they need books to develop their models,” said Authors Guild CEO Mary Rasenberger.

Keith Kupferschmid, president of the Copyright Alliance, said the settlement amount not only represented victory, it proved what authors had been saying all along: “that AI companies can afford to compensate copyright owners for their works without it undermining their ability to continue to innovate and compete.”

Anthropic, now valued at $310 billion following a fresh $13bn (NZ$22) funding round, said the settlement resolves “legacy claims” as it continues developing “safe AI systems.”