Anthropic agrees to pay $1.5 bn to settle pirated books lawsuit with authors

Anthropic PBC has consented to a settlement amounting to at least $1.5 billion, plus interest, to resolve a class action lawsuit initiated by authors. The lawsuit accused the AI startup of unlawfully downloading millions of pirated books to enhance its language models. Filed in a San Francisco federal court, this proposed settlement stands as one of the largest in the realm of intellectual property and artificial intelligence. With a trial originally set for December, plaintiffs were pursuing damages that could have escalated into trillions, posing a serious threat to the company's viability. Remarkably, this agreement addresses claims from approximately seven million authors, with a projected compensation of about $3,000 per book for an initial batch of 500,000 works. If additional claims prove valid, the total payout could increase. In a further commitment, Anthropic has agreed to eliminate any disputed data it was accused of wrongfully acquiring. Justin Nelson, an attorney representing the authors, labeled this settlement as unprecedented, stating, "This settlement sends a powerful message to AI companies and creators alike that taking copyrighted works from pirate websites is wrong." Despite the considerable financial implications, legal experts believe that Anthropic has avoided an even more severe outcome. Tod Cohen, a partner at Manatt, Phelps & Phillips, noted, "It is less than one percent of the company’s valuation. It seems like a win for Anthropic that it is not required to dismantle its models, just delete the data." Established in 2021, Anthropic has quickly emerged as a key competitor to OpenAI, securing $13 billion in investments and reaching an estimated annual revenue run-rate of $5 billion, although it has yet to achieve profitability due to the high costs associated with developing its flagship Claude AI systems. This settlement is among the first to be concluded in a series of copyright lawsuits filed against AI firms, including OpenAI, Meta, and Midjourney, regarding their use of proprietary online content. Unlike some competitors who have proactively entered licensing agreements with publishers to circumvent litigation, Anthropic has not disclosed any such arrangements. The company still faces ongoing legal challenges from music publishers alleging unauthorized copying of lyrics, as well as from Reddit, which claims that its platform content was used without consent for training purposes. Legal analysts suggest that this San Francisco settlement might establish a precedent within the industry, although it remains uncertain whether it will lead to a broader wave of settlements. Chad Hummel, an attorney at McKool Smith, remarked, "This is a landmark event, the first major settlement in a case against a generative AI company. It could prove industry guiding." A hearing regarding the proposed settlement is scheduled for September 8.