Anthropic has reached a settlement of $1.5bn in a class-action lawsuit filed by a group of authors who accused the company of using their books without permission to train its AI chatbot, Claude.
The settlement was submitted to US District Judge William Alsup in San Francisco, with both parties requesting court approval.
The authors, who initiated the lawsuit last year, claimed that Anthropic unlawfully incorporated millions of pirated books into its training data.
In a court filing, the plaintiffs noted, “If approved, this landmark settlement will be the largest publicly reported copyright recovery in history, larger than any other copyright class action settlement or any individual copyright case litigated to final judgment.”
This settlement is the first in a series of lawsuits against technology companies, including OpenAI, Microsoft, and Meta Platforms, regarding the use of copyrighted materials for training generative AI systems.
As part of the settlement, Anthropic has committed to destroying the downloaded copies of the books that the authors claimed were pirated.
However, the company may still face infringement claims related to outputs generated by its AI models.
In a statement, Anthropic reiterated its focus on developing AI systems, but the settlement does not involve an admission of liability.
The class action was initiated last year by writers Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who contended that Anthropic, supported by Amazon and Alphabet, unlawfully employed millions of pirated books to train Claude to respond to user prompts.
Their claims are part of a wave of lawsuits from authors, media organisations, visual artists, and others alleging that tech companies have misappropriated their works for AI training purposes.
While the companies involved have argued that their systems engage in fair use of copyrighted content to generate new and transformative material, Judge Alsup ruled in June that Anthropic had indeed made fair use of the authors’ works for training Claude.
However, he also determined that the company infringed on their rights by storing more than 7 million pirated books in a “central library” that was not exclusively designated for that purpose.
A trial was set to commence in December to ascertain the extent of damages owed by Anthropic for the alleged piracy, with potential liabilities reaching into the hundreds of billions of dollars.
The critical fair-use issue continues to be a point of contention in other ongoing AI copyright disputes.
Following Alsup’s ruling, another San Francisco judge presiding over a similar case against Meta indicated that using copyrighted material without permission for AI training could be deemed unlawful in “many circumstances.”
