Federal judge grants final approval to Anthropic's $1.5 billion AI copyright settlement

A federal judge in San Francisco granted final approval on July 20, 2026 to Anthropic's $1.5 billion settlement with authors over pirated books used to train Claude, the largest known recovery in the history of U.S. copyright law and the first major settlement among the wave of AI copyright suits working through U.S. courts.

What's new

U.S. District Judge Araceli Martinez-Olguin, who took over the case Bartz v. Anthropic after Judge William Alsup's earlier rulings, signed off on the settlement despite objections from some authors who argued the payout was too small. The deal covers an estimated 500,000 works and pays roughly $3,000 per book to authors and publishers who hold rights to them; more than 91% of eligible claimants had already filed for their share by the time of approval. As part of the settlement, Anthropic is required to destroy the pirated copies of the books it had stored. The court separately awarded plaintiffs' attorneys more than $101 million in fees, below the $187.5 million they had requested.

The underlying case traces back to a 2024 suit brought by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who accused Anthropic of large-scale copyright infringement by downloading and commercially exploiting books obtained from pirated datasets, including sites like Library Genesis and Pirate Library Mirror. Judge Alsup had earlier ruled that training Claude on copyrighted text can qualify as fair use, but found that Anthropic's practice of downloading and retaining more than 7 million pirated books in a central library was independently unlawful, separate from whether those books were ultimately used in training.

Context

The settlement was first proposed in September 2025, when Anthropic agreed to pay $1.5 billion to resolve the claims, but Judge Alsup initially withheld approval and raised concerns about the adequacy of the deal before a hearing process addressed his objections. Anthropic is one of several AI developers facing copyright suits from authors, publishers, and news organizations over training data; this settlement is the first of those cases to reach a final, court-approved resolution rather than continuing through litigation.

Some authors and publishers opted out of the class settlement and have filed separate, still-pending lawsuits against Anthropic, meaning the underlying legal questions about AI training data are not fully resolved even though this particular case is closed.

Why it matters

As the first major AI copyright case to settle rather than proceed to trial or a definitive appellate ruling, this outcome sets a financial benchmark — roughly $3,000 per infringed work — that plaintiffs' attorneys in other pending AI copyright suits are likely to point to in negotiations with other model developers. It also draws a sharper line for AI labs generally: courts may tolerate training on copyrighted text as fair use, but sourcing that text through piracy carries separate, substantial legal exposure regardless of how the fair-use question is ultimately resolved. For Anthropic specifically, closing out its largest pending legal liability removes a significant overhang as the company continues to raise capital and expand commercially.