A U.S. federal judge has blocked Anthropic’s proposed $1.5 billion settlement with authors who accused the AI company of training its Claude models on pirated books, citing serious concerns about transparency and potential unfairness to writers. The ruling by Judge William Alsup on September 13th allows the landmark copyright case to proceed while emphasizing the need for fair compensation and proper notice to approximately 465,000 affected authors.
The Massive Settlement Put on Hold
Anthropic had agreed to pay $1.5 billion to resolve a class-action lawsuit filed by thriller writer Andrea Bartz and nonfiction authors Charles Graeber and Kirk Wallace Johnson, who alleged the company used pirated copies of their works to train its AI chatbot Claude without permission.
Under the original settlement terms, authors and publishers would receive approximately $3,000 per work for an estimated 500,000 books included in the agreement. Legal experts called it “the largest copyright recovery in history” and potentially the first major payout in the AI era.
However, Judge William Alsup expressed serious reservations during Monday’s hearing, stating he had “an uneasy feeling about hangers-on with all this money on the table.” The judge worried that class action attorneys might forge a private deal that could be imposed on authors against their will.
Judge’s Key Concerns About the Deal
Judge Alsup identified several critical problems with the proposed settlement that needed addressing before approval:
Unclear Claims Process: The judge demanded more details about how authors would actually claim their compensation, expressing concern about administrative burdens that could make the process unmanageable for writers.
Incomplete Author Lists: Alsup required comprehensive lists of all works and authors affected by the settlement by September 15th, emphasizing that “class members must receive very good notice” to guarantee their awareness of the case.
Risk of Duplicate Lawsuits: The judge flagged concerns about whether Anthropic might face additional unexpected lawsuits if payouts weren’t clearly defined, demanding a definitive figure of affected works.
Transparency Issues: Alsup criticized the lack of transparency in the settlement negotiations, suggesting that authors might be pressured into unfair terms without a proper understanding.
The Original Copyright Infringement Case
The lawsuit began when authors accused Anthropic of using over 7 million pirated books from shadow libraries, including Books3, Library Genesis (LibGen), and Pirate Library Mirror, to train its Claude AI models without authorization.
In June 2025, Judge Alsup delivered a mixed ruling that became crucial to the case. He determined that training AI on legally purchased books qualified as “fair use” under copyright law, calling it “among the most transformative we will see in our lifetimes”.
However, the judge found that Anthropic’s use of pirated materials was “inherently, irredeemably infringing” regardless of the purpose. This ruling left Anthropic vulnerable to potentially billions in damages for the illegally obtained books, prompting the settlement negotiations.
Industry Reactions and Implications
Publishers and author groups expressed frustration with the judge’s decision to halt the settlement. Maria Pallante, president of the Association of American Publishers, argued that Alsup “has shown a lack of understanding of how the publishing industry operates”.
Authors Guild CEO Mary Rasenberger said her organization was “shocked by the court’s offhand suggestion” that groups were “working behind the scenes in ways that could pressure authors,” defending the settlement as fair compensation.
However, some legal experts supported the judge’s caution. Trial lawyer Chad Hummel told The New York Times that “this is massive” and “will cause generative AI companies to sit up and take notice,” suggesting the case could reshape how AI companies negotiate with content creators.
What Happens Next
Both parties must provide additional information by September 22nd to address the judge’s concerns about the settlement structure and claims process. The court has scheduled further hearings to review these submissions.
If the settlement ultimately fails, Anthropic faces a high-stakes trial originally scheduled for December that could result in even higher damages. Legal analyst William Long warned that “we were facing a strong likelihood of several billion dollars in damages, which could have severely harmed or even led to the closure of Anthropic”.
Broader Impact on the AI Industry
This case represents the first major settlement attempt in a wave of lawsuits targeting tech companies over alleged misuse of copyrighted material for AI training. Similar cases are pending against Apple, OpenAI, Meta, and other major AI developers.
The outcome could establish precedents for how AI companies must compensate creators for using their works, potentially requiring licensing fees or similar arrangements going forward. As one expert noted, this could “pave the way for these companies to pay rights holders through licensing fees”.
The settlement, if eventually approved, would not give Anthropic or any other AI company permission to use pirated books going forward – it only resolves Anthropic’s liability for past alleged infringement.
Judge Alsup’s decision to halt the settlement demonstrates the courts’ determination to ensure fair treatment for creators in the rapidly evolving landscape of AI development and copyright law
