Federal judge approves Anthropic's $1.5B copyright settlement, but fair-use question remains unresolved industrywide
Anthropic's landmark settlement closes a piracy case but leaves AI training copyright rules fragmented across competing court decisions.
Last verified:
Anthropic has secured final approval for its $1.5 billion copyright settlement, according to TechCrunch AI, with Judge Araceli Martinez-Olguin signing off on July 21, 2026. The settlement resolves a class-action lawsuit brought by authors and publishers over the unauthorized use of copyrighted books in the AI lab’s training dataset. However, the case’s resolution leaves the broader question of AI training and copyright—the question every AI company faces—legally unsettled across the U.S. court system.
The settlement structure and payout
The settlement will distribute $3,000 per work across an estimated 500,000 copyrighted titles, with payments flowing to authors and publishers who retain rights to those works. TechCrunch reports this is the largest copyright settlement in U.S. history. Judge William Alsup, who issued preliminary approval in 2025, had ruled that Anthropic illegally downloaded and stored millions of books from pirate sites including Library Genesis and Pirate Library Mirror—a method distinct from the company’s legal practice of purchasing and scanning books.
Fair use approved, but piracy condemned
The settlement’s legal logic reveals a split decision that satisfies neither authors nor the AI industry. According to TechCrunch, Alsup ruled that training an AI model on copyrighted text itself constitutes fair use—a decision widely interpreted as a turning point favoring AI labs. However, Alsup distinguished between the legality of using copyrighted material for training and the legality of obtaining it through piracy. He found Anthropic’s reliance on pirated sources illegal on its own terms, separate from the fair-use question. Anthropic agreed to settle rather than face a jury trial on damages.
Why other lawsuits remain unaffected
Because Anthropic settled rather than appealing, Alsup’s fair-use ruling never reached appellate court and carries no binding precedent outside the Northern District of California. As TechCrunch notes, other judges remain free to reach different conclusions. OpenAI, Google, Meta, and Midjourney face active copyright lawsuits with no coordinated resolution in sight. TechCrunch reports that last week Hachette, Cengage, Elsevier, author Scott Turow, and S.C.R.I.B.E. filed a new class action against Google alleging unauthorized use of copyrighted works to train Gemini.
Why This Matters
For Anthropic, the settlement closes a legal exposure and allows the company to move forward without ongoing trial risk. For the broader AI industry, however, the ruling offers only partial guidance. The fair-use determination is encouraging to AI labs but remains a single district court’s opinion, not binding law. Individual authors and publishers view the $1.5 billion settlement skeptically—the per-work payout of $3,000 is modest relative to claimed damages, and the settlement includes no injunction against future training on copyrighted material. For AI teams selecting training data sources, the decision reinforces that purchasing legitimate licenses is safer than relying on pirate archives, but it does not resolve whether fair use permits training on any copyrighted material at all. That fragmented legal landscape will persist until an appellate court—or Congress—provides industry-wide clarity.
Frequently Asked Questions
Did Anthropic win or lose this case?
Anthropic won on the core legal question—the judge ruled that training AI models on copyrighted text counts as fair use. However, the company settled over how it obtained the books: downloading from pirate sites like Library Genesis was found illegal, even though scanning purchased books was not.
Why does this settlement not set precedent for the whole AI industry?
Because Anthropic settled rather than appeal, the case never reached a higher court. Judge Alsup's ruling is binding only within the Northern District of California. Other judges in other jurisdictions remain free to rule differently, which is why copyright lawsuits against OpenAI, Google, Meta, and Midjourney are still proceeding.
How much will individual authors receive?
According to the settlement terms, each copyrighted work receives $3,000, distributed among the authors and publishers who hold rights to the approximately 500,000 works in Anthropic's training dataset.