Anthropic faces cost of AI training data piracy, agrees to pay $1.5B to 7M authors
Anthropic PBC has agreed to pay at least $1.5 billion plus interest to resolve a class action lawsuit brought by authors who accused the artificial intelligence start-up of illegally downloading millions of pirated books to train its language models.
One of the largest AI-related settlements
The proposed settlement, filed on Friday in San Francisco federal court, is one of the largest agreements yet over intellectual property and artificial intelligence. A trial in the case had been scheduled for December, with the plaintiffs seeking damages that could have run into the trillions and threatened the company’s survival.
Compensation for millions of authors
Notably, the deal covers claims from as many as seven million authors, with compensation estimated at about $3,000 per book for an initial pool of 500,000 works. Should additional claims be validated, the total payout could rise. Anthropic has also committed to deleting any disputed data it was accused of unlawfully acquiring.
Lawyers representing the authors described the accord as unprecedented. “This settlement sends a powerful message to AI companies and creators alike that taking copyrighted works from pirate websites is wrong,” said Justin Nelson of Susman Godfrey.
While the agreement represents a significant financial hit, legal experts suggested Anthropic avoided a far more damaging outcome. “It is less than one per cent of the company’s valuation,” said Tod Cohen, partner at Manatt, Phelps & Phillips. “It seems like a win for Anthropic that it is not required to dismantle its models, just delete the data.”
Founded in 2021, Anthropic has rapidly become one of the most closely watched challengers to OpenAI. The company has attracted $13 billion in investment and achieved an estimated $5 billion annualised revenue run-rate, though it remains unprofitable due to the costs of developing its flagship Claude AI systems.
The settlement is among the first to be reached in a wave of copyright lawsuits filed against AI firms, including OpenAI, Meta and Midjourney, over their use of proprietary online content. Some competitors have pre-emptively struck licensing deals with publishers to avoid litigation; Anthropic has not disclosed any such agreements.
Ongoing legal battles
The company still faces ongoing cases from music publishers, who allege unlawful copying of lyrics, and from Reddit, which claims its platform content was used without permission for training.
Legal observers say the San Francisco deal may set a precedent for the industry, though it remains unclear whether it will trigger a wider series of settlements. “This is a landmark event, the first major settlement in a case against a generative AI company,” said attorney Chad Hummel of McKool Smith. “It could prove industry guiding.”
A hearing on the proposed settlement is scheduled for 8 September.
(With inputs from Bloomberg)
