Case summary
- Plaintiffs
- Andrea Bartz, Charles Graeber, Kirk Wallace Johnson — representing ~500,000-member class
- Defendant
- Anthropic PBC
- Court
- U.S. District Court, Northern District of California
- Case number
- 3:24-cv-05417-WHA
- Filed
- August 19, 2024
- Judge
- Hon. William Alsup
- Claim types
- Direct copyright infringement (willful) — pirated-book training data
- Settlement
- $1.5 billion — ~$3,000 per work to 500,000 authors
- Status
- Preliminary approval granted; final fairness hearing April 2026
- Class counsel
- Susman Godfrey LLP · Lieff Cabraser Heimann & Bernstein LLP
TL;DR
Three authors sued Anthropic in August 2024 alleging it trained Claude on pirated books downloaded from shadow libraries like Library Genesis and Pirate Library Mirror. In June 2025, Judge Alsup issued a bifurcated ruling: training on legitimately acquired books was fair use, but Anthropic's downloading and retention of pirated copies was not protected. With a willful-infringement jury trial looming and potential statutory damages up to $150,000 per work across hundreds of thousands of books, Anthropic settled in August 2025 for $1.5 billion — the largest copyright settlement in U.S. history.
Anthropic will pay approximately $3,000 per work to a ~500,000-author class. Class counsel has requested $300 million in fees (20% of the fund). The opt-out deadline passed January 15, 2026; final fairness hearing is scheduled for April 2026.
Why it matters
- Sets the damages floor for every pending books case. $3,000/work × class size is now the benchmark any plaintiff class will anchor to.
- Bifurcates the fair-use question in a way that matters: training on licensed material may be fair use; training on pirated material is not. Every AI lab is auditing its corpora accordingly.
- $300M in attorney fees incentivizes plaintiffs' bar to file more AI cases. Expect more, not fewer.
- Anthropic must destroy the pirated datasets and certify they are not present in any commercially deployed model. This is a precedent for injunctive relief targeting deployed LLMs.
Timeline of events
- Aug 19, 2024Class action complaint filedBartz, Graeber, and Johnson file in NDCA alleging Claude was trained on pirated copies of their books.
- Feb 2025Class certification briefingPlaintiffs move to certify a class covering all authors whose books appear in Books3 and LibGen datasets.
- Jun 2025Bifurcated fair-use ruling (Alsup, J.)Court holds training is fair use as to legitimately acquired copies; not fair use as to pirated copies. Willful infringement damages survive for trial.
- Aug 2025$1.5B settlement announcedParties announce the largest copyright settlement in U.S. history.
- Sep 2025Preliminary approval grantedJudge Alsup grants preliminary approval; notice program commences.
- Dec 04, 2025Class counsel requests $300M in feesSusman Godfrey & Lieff Cabraser file fee petition: 20% of settlement, ~26,000 hours worked.
- Jan 15, 2026Opt-out deadlineDeadline for class members to opt out and pursue individual actions.
- Apr 2026Final fairness hearingJudge Alsup to review settlement fairness and objections from authors opposing the deal.
Rulings & outcomes
Bifurcated fair-use order
Jun 2025 · Alsup, J.Holding: Anthropic's method for training AI models on legitimately acquired books qualifies as fair use. However, Anthropic's acquisition and retention of pirated copies from shadow libraries was not fair use and is actionable willful infringement.
Anthropic could have lawfully purchased the books … but chose to "steal" them to circumvent what the company's CEO referred to as "legal/practice/business slog."
Preliminary approval of $1.5B settlement
Sep 2025 · Alsup, J.Holding: Court preliminarily approves the class-wide settlement. Settlement requires Anthropic to destroy pirated datasets, pay ~$3,000 per work, and allow class members a 120-day opt-out window.
Money at stake
Press coverage
- New York TimesAnthropic agrees to pay $1.5 billion to settle lawsuit with book authors
- ReutersAuthors' lawyers in $1.5 billion Anthropic settlement seek $300 million
- Kluwer Copyright BlogThe Bartz v. Anthropic settlement: understanding America's largest copyright settlement
- Authors AllianceAI class action litigation update (books), early 2026
FAQ
Is the settlement final?
Preliminary approval was granted in September 2025. Final approval requires the April 2026 fairness hearing before Judge Alsup. Objections from class members opposing the agreement will be considered at that hearing.
How much does each author get?
Approximately $3,000 per work. Authors with multiple works will receive correspondingly more. For books with multiple authors, the author's 50% share is divided among them, while publishers receive the remaining 50% by default (adjustable by contract).
Does the settlement mean training on books is legal now?
No. Judge Alsup's June 2025 order held training on legitimately acquired copies may be fair use, but using pirated copies is not. The settlement resolves liability for pirated acquisition specifically.
Are Claude models affected?
Anthropic represents that the pirated datasets were not used to train commercially deployed models. As part of the settlement, Anthropic must destroy the pirated materials.