AI start-up Anthropic has agreed to pay $1.5bn to settle a copyright lawsuit over its use of pirated texts, setting a precedent for tech companies facing intellectual property cases from authors and publishers.
The settlement, which must be approved by the San Francisco federal judge overseeing the case, would be “the largest publicly reported copyright recovery in history”, according to a court filing on Friday.
The class action suit was brought by authors who claimed Anthropic had downloaded 465,000 books and other texts from “pirated websites” including Library Genesis and Pirate Library Mirror, which it then used to train its large language models.
Failure to reach an agreement would have led to a trial, with the prospect of damages of up to $1tn. That would have bankrupted Anthropic, a four-year-old start-up backed by Amazon and Google that was recently valued at $170bn.
The case is among several facing artificial intelligence companies, including OpenAI and Meta, alleging they have improperly used copyrighted works to train their models.
The results will help determine how authors are compensated for the use of their works and could have significant ramifications for how AI companies train their models and the costs of developing them.
Mary Rasenberger, chief executive of the Authors Guild, said Friday’s settlement sent a “strong message to the AI industry that there are serious consequences when they pirate authors’ works to train their AI, robbing those least able to afford it”.
Anthropic and other AI companies have claimed that training models on copyrighted books is fair use, arguing that their models transform the original work into something with a new meaning.
In June, the California district court ruled that Anthropic’s use of some copyrighted works in such a way was fair. But it determined that storing pirated works was “inherently, irredeemably infringing”, teeing up Friday’s settlement.
Friday’s ruling also means Anthropic will have to destroy the datasets it had downloaded from Library Genesis and Pirate Library Mirror.
“In June, the district court issued a landmark ruling on AI development and copyright law, finding that Anthropic’s approach to training AI models constitutes fair use. Today’s settlement, if approved, will resolve the plaintiffs’ remaining legacy claims,” said Anthropic’s deputy general counsel Aparna Sridhar in a statement.
“We remain committed to developing safe AI systems that help people and organisations extend their capabilities, advance scientific discovery and solve complex problems,” she added.