A federal judge has given final approval to Anthropic's $1.5 billion settlement of a class action copyright lawsuit, formally closing the largest copyright settlement in U.S. history. Judge Araceli Martinez-Olguin of the U.S. District Court for the Northern District of California signed off on the final approval on Monday, July 20, 2026, as first reported by Reuters (Reuters).

The case, formally known as Bartz v. Anthropic, centered on Anthropic's construction of its AI training library from two sources: books it purchased and scanned, and books downloaded from pirate sites including Library Genesis and Pirate Library Mirror. Anthropic reportedly downloaded approximately 7 million copies of books in total, of which roughly 500,000 titles are covered by the settlement (Authors Guild).

Judge William Alsup, now retired, issued the preliminary approval and had previously ruled on the two components of Anthropic's data acquisition separately. Alsup found that training an AI model on copyrighted text counts as fair use. However, he ruled that Anthropic's downloading of books from pirate sites was illegal. Anthropic settled to avoid a trial and potential jury-awarded damages (TechCrunch).

The settlement payout works out to $3,000 per work across an estimated 500,000 works, to be shared among the authors and publishers who hold rights to them. Anthropic agreed to pay the $1.5 billion regardless of how many individual rightsholders opt out of the settlement, and the payment is structured over time in several installments (Authors Alliance; City Law Forum).

The path to final approval was not straightforward. After the settlement was announced in August 2025, a federal judge in San Francisco initially declined to grant approval (Reuters). Preliminary approval came on Thursday, September 25, 2025 (Reuters). Martinez-Olguin later delayed final approval at one point, seeking more details before ultimately granting it (Storyboard18). The lead attorneys for the plaintiff class also reduced their fee request in March 2026 after receiving pushback (Reuters).

A critical structural consequence of the settlement is that Alsup's fair-use ruling will never reach an appeals court and therefore does not become binding precedent. The distinction matters: training on copyrighted text was found to be fair use at the district court level, but that finding carries no binding weight outside the Northern District of California. Other AI developers facing similar lawsuits are not bound by it, and plaintiffs in those cases are free to argue the opposite.

The broader litigation landscape remains active. Ongoing copyright lawsuits over AI training on copyrighted works involve companies including Google, Meta, Midjourney, and OpenAI (TechCrunch). In the week of July 14, 2026, publishers and authors including Hachette, Cengage, Elsevier, author Scott Turow, and S.C.R.I.B.E. filed a class action lawsuit against Google over accusations that it used their copyrighted works to train its AI platform Gemini (TechCrunch).

The Anthropic settlement resolves one case but leaves the central legal question unsettled. Whether training a commercial AI model on copyrighted text constitutes fair use under U.S. law remains unresolved at the appellate level. Anthropic's decision to settle means the company pays $1.5 billion for the piracy component of its data pipeline while preserving a favorable district-court fair-use finding that no other court is obligated to follow.

For the AI industry, the settlement establishes a rough price point: roughly $3,000 per copyrighted work for the act of illicitly downloading books from pirate sources. Whether that figure becomes a de facto benchmark depends on outcomes in the pending cases against Google, Meta, OpenAI, and others, where the facts around data acquisition and use may differ. The fair-use question, which is the one developers most need answered, remains open.