Anthropic has reached a preliminary settlement in a class action lawsuit brought by a group of prominent authors, resolving one of the most closely watched copyright disputes in the artificial intelligence sector. The agreement, which is expected to be finalised on 3 September 2025, allows the AI company to avoid potentially catastrophic financial damages while offering compensation to writers whose works were used without permission.
The lawsuit was filed in 2024 by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who accused Anthropic of training its Claude AI systems on copyrighted books sourced from so-called shadow libraries such as LibGen and Pirate Library Mirror
In June 2025, U.S. District Judge William Alsup ruled that training AI on lawfully purchased or licensed works qualified as fair use, but he also determined that Anthropic’s reliance on pirated material constituted a clear copyright violation. That finding opened the way for a class action trial that had been set for December.
Under U.S. copyright law, statutory damages start at 750 dollars (R13,900) per infringed work and can rise to 150,000 dollars (R2.8 million) for each instance of wilful infringement. With Anthropic’s dataset estimated at around 7 million works, its potential liability was measured in the billions and described by some scholars as “doomsday-level damages”.
Speaking on behalf of the plaintiffs, attorney Justin Nelson called the settlement “historic” and emphasised that it will benefit all members of the class. Many affected authors were only recently informed of their eligibility through notices circulated by the Authors Guild.
Implications for the publishing industry and AI companies
The deal has been welcomed by both the Authors Guild and the Association of American Publishers, who praised it as an important affirmation of copyright protections in the age of generative AI. For Anthropic, the agreement prevents a December trial that could have crippled the company, which is still operating at a loss and facing other high-profile lawsuits.
The firm is currently defending itself against claims by major record labels, who allege unlawful use of song lyrics in training data, and against legal action over the scraping of online content from platforms such as Reddit.
While this settlement resolves the authors’ case, it does not create a binding legal precedent. Ongoing litigation against OpenAI, Microsoft, and Meta means that the legal boundaries of AI training remain unsettled and will continue to evolve in the months ahead.
What happens next
Judge Alsup is expected to review motions for preliminary approval of the settlement by 5 September, with a hearing scheduled for 8 September. If the deal is approved, the Authors Guild will provide further details to affected authors regarding compensation. If not, the December class action trial could still proceed.
The outcome of this case will be closely monitored as a bellwether for the broader clash between innovation in artificial intelligence and the rights of human creators.
