Anthropic's $1.5B copyright settlement gets final court approval

On Monday, July 21, 2026, Judge Araceli Martinez-Olguin of the U.S. District Court for the Northern District of California gave final approval to Anthropic's $1.5 billion（約2400億円） settlement with authors and book publishers over copyright infringement claims. The settlement will distribute $3,000 per work across an estimated 500,000 works, with payments shared among authors and publishers who hold rights to the copyrighted materials.

The path to final approval began when Judge William Alsup issued a preliminary ruling last year after determining that Anthropic had illegally downloaded and stored millions of copyrighted books. Critically, Alsup split the copyright question into two parts: on the core legal issue, he sided with Anthropic, ruling that training an AI model on copyrighted text constitutes fair use—a finding the AI industry has viewed as a major victory. However, Alsup's decision to favor the fair-use defense did not absolve Anthropic of liability for how it obtained the books. The company had built its training library from two sources: books it purchased and scanned, and books it downloaded from pirate sites including Library Genesis and Pirate Library Mirror. Alsup found the piracy method illegal on its own and indicated that the question could proceed to trial. Rather than face a jury trial and the risk of potentially larger damages, Anthropic agreed to settle.

Despite being hailed as the largest copyright settlement in U.S. history, many authors and creators view the outcome as incomplete. The settlement does not establish binding legal precedent because it is a district court decision and Anthropic chose settlement over appeal. This means other judges in other cases remain free to reach their own conclusions based on their own facts. As evidence of this fragmentation, a string of copyright lawsuits continues against other AI companies. Last week alone, a group of publishers and authors—including Hachette, Cengage, Elsevier, author Scott Turow, and S.C.R.I.B.E.—filed a class action against Google alleging the company used their copyrighted works to train Gemini. Similar litigation remains pending against Meta, Midjourney, and OpenAI, leaving the legal status of AI training on copyrighted materials uncertain across the industry.

Anthropic's settlement marks a complex turning point in the AI industry's copyright battles. While Judge Alsup's preliminary ruling that training AI on copyrighted text qualifies as fair use was widely interpreted as a major win for the AI industry, Anthropic's path to settlement reveals a critical distinction: the company's method of obtaining the books mattered as much as the use itself. By downloading millions of copyrighted works from pirate sites rather than obtaining them through legitimate channels, Anthropic exposed itself to liability that went beyond the "fair use" question. The company chose to settle before a jury trial rather than risk much larger damages—suggesting that despite the favorable fair-use ruling, the illegality of the acquisition method was significant enough to warrant the landmark $1.5 billion（約2400億円） payout.

The settlement's status as the largest in U.S. copyright law history does not translate to industry-wide legal clarity. Precisely because Anthropic settled and avoided appeal, Alsup's fair-use ruling remains a single district court decision with no binding power over other courts or cases. This leaves the broader legal landscape fragmented: Google, Meta, Midjourney, and OpenAI each face ongoing copyright suits, and judges in those cases are free to interpret fair use differently, especially if the facts differ (for example, if an AI company obtained training data through legitimate licensing rather than downloading from pirate sites). Last week's fresh class action against Google by major publishers and authors signals that the copyright question is far from resolved.