Anthropic informed a federal judge in San Francisco on Friday about its agreement to pay $1.5 billion to settle a class-action lawsuit filed by a group of authors. The authors accused the artificial intelligence company of utilizing pirated copies of their books to train its AI chatbot, Claude, without obtaining permission.
Both Anthropic and the plaintiffs requested U.S. District Judge William Alsup to approve the settlement after revealing the agreement in August, initially without disclosing the terms or the amount involved. The plaintiffs emphasized in the court filing that if approved, this settlement would be the largest publicly reported copyright recovery to date, surpassing any other copyright class action settlement or individual copyright case taken to final judgment.
This proposed settlement signifies the initial resolution in a series of legal actions against technology firms such as OpenAI, Microsoft, and Meta Platforms for their use of copyrighted content in training generative AI systems. As part of the settlement, Anthropic committed to eradicating downloaded book copies acquired from piracy sites like LibGen and PiLiMi (Pirate Library Mirror). However, the company could still potentially face infringement claims related to content generated by its AI models.
In a statement, Anthropic stressed its dedication to developing secure AI systems that aid individuals and organizations in enhancing their capabilities, fostering scientific discoveries, and solving intricate problems. Notably, the agreement does not include an admission of liability.
The Authors Guild reported that around 500,000 works are covered under this settlement, indicating that each author stands to receive an estimated $3,000. Authors Guild CEO Mary Rasenberger hailed this historic settlement as a crucial acknowledgment that AI companies cannot unlawfully appropriate authors’ creative works for developing quality LLMs without compensating the creators.
While approximately seven million books were reportedly downloaded by Anthropic from piracy platforms, the class-action lawsuit only covers about 500,000 works. Noteworthy writers Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson initiated the lawsuit against Anthropic last year, alleging the unauthorized utilization of millions of pirated books by the company to educate its AI assistant, Claude, on human interactions.
The authors’ claims echoed those of numerous other litigants, including writers, news outlets, and visual artists, who alleged that tech companies misappropriated their works for AI training purposes. Alsup previously ruled that Anthropic did make fair use of the books to train Claude but breached the authors’ rights by storing over seven million pirated books in a central library not necessarily intended for that specific purpose.
A trial was slated to commence in December to ascertain the compensation owed by Anthropic for the alleged piracy, with potential damages estimated in the billions of dollars. The ongoing debate over the pivotal fair-use issue persists in various other AI copyright litigations. Vancouver author J.B. MacKinnon recently initiated class-action lawsuits against NVIDIA, Meta, Anthropic, and Databricks Inc. in the B.C. Supreme Court, alleging the unauthorized use of his and other Canadian authors’ works for AI training purposes.
Another San Francisco judge handling a related lawsuit against Meta concurred shortly after Alsup’s ruling that utilizing copyrighted material without authorization for AI training could be deemed unlawful in numerous scenarios.