🔍 Read the full analysis: The Claims About Pirated Books In Claude’s Training, Explained on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
A 2025 federal court ruling said Anthropic’s use of books to train Claude was fair use, but found that keeping more than seven million pirated book copies was not. Anthropic later agreed to a $1.5 billion settlement covering about 500,000 books, without admitting wrongdoing; new authors’ lawsuits remain unresolved.
Anthropic’s use of books to train Claude was ruled fair use in a June 2025 copyright case, but a federal judge separately found that the company could not justify downloading and keeping more than seven million pirated book copies. The distinction matters: the ruling addressed copying for a permanent digital library as unlawful, while claims in newer lawsuits about authors’ books and models remain unadjudicated. Anthropic later agreed to a $1.5 billion settlement over roughly 500,000 books, while denying wrongdoing.
The ruling came in Bartz v. Anthropic, a case brought by three authors in 2024. On June 23, 2025, Judge William Alsup ruled that training a language model on books could be fair use, describing Anthropic’s training as an effort “to turn a hard corner and create something different.” He also found that Anthropic’s purchase and scanning of millions of used print books was fair use. Those findings did not extend to the company’s acquisition and retention of pirated files.
Alsup’s order described 196,640 books from Books3, at least five million copies from Library Genesis, or LibGen, and at least two million from Pirate Library Mirror, or PiLiMi. He wrote that every fair-use factor weighed against Anthropic for the pirated copies and that buying a copy later would not erase liability for the earlier theft. The ruling was a pretrial decision, not a jury verdict on damages; the parties settled before a jury decided the remaining issues.
The settlement, finally approved in July 2026, provides $1.5 billion for owners of about 500,000 registered books downloaded from LibGen or PiLiMi—about $3,000 per book before fees. It requires Anthropic to destroy the downloaded books and copies, subject to legal-hold obligations. The settlement covers specified past conduct through August 25, 2025; it does not release claims about Claude’s outputs. Anthropic disputes wrongdoing and represented to the settlement court that neither dataset, nor portions of them, appeared in the training corpus of any commercially released model.
From download to IPO
Five years, one ruling, no trial
The piracy happened in 2021 and 2022. The only ruling came in 2025. The new lawsuits are years away from a verdict.
- Jun 2021 · Jul 2022DownloadsAt least 5 million copies from LibGen (2021), at least 2 million from PiLiMi (2022); 196,640 from Books3 earlier.
- Aug 2024Bartz v. Anthropic filedThree authors sue Anthropic, on behalf of a class of book owners.
- 23 Jun 2025Judge Alsup's orderTraining on books: fair use. Scanning bought books: fair use. Pirated library copies: not fair use, to trial.
- Sep 2025$1.5 billion settlementNo trial. Anthropic denies wrongdoing; release covers conduct up to 25 Aug 2025 only.
- 13 + 15 May 2026Opt-out suitsCruz (28 plaintiffs) and Kwon (incl. Chang-rae Lee) join Cambronne and two publisher suits.
- Jul 2026Final approvalSettlement approved. Anthropic: no opt-out deal "above the Bartz claim value".
- 24 Sep 2026Motion to dismissAnthropic: claims too late, no infringing output shown, many works already released.
- 26–27 Oct 2026NowS-1 and the headlineIPO filing cites Bartz as a risk; The Stanford Daily names Stanford authors.
- 17 Dec 2026ScheduledHearingCourt hears the motion to dismiss.
- May 2027 · Jan 2028ScheduledDiscovery, then trial settingFact discovery closes 13 May 2027; trial-setting conference 11 Jan 2028.
Sources: Bartz v. Anthropic, order of 23 June 2025; settlement FAQ; joint case statement, 27 Aug 2026; Anthropic letter, 27 Jul 2026; case management order, 14 Sep 2026; Anthropic S-1, 26 Oct 2026; The Stanford Daily, 27 Oct 2026. Scheduled dates can move.
Powered by Thorsten Meyer AI
What the Settlement Covers
The distinction between training on books and obtaining books through piracy has practical consequences for authors, AI companies and copyright law. The 2025 ruling did not establish that all AI training on copyrighted material is lawful, nor did it grant Anthropic a general licence to use books. It found that particular training and scanning practices in this case qualified as fair use, while rejecting the justification for retaining pirated copies.
The settlement offers compensation to eligible rights holders without a jury deciding damages or the broader legal questions. Its scope is limited: it addresses registered books in the identified collections and past conduct, not every author or every potential claim. For readers and creators, it is a substantial resolution of one part of the dispute, but not a final answer about copyright rules for generative AI.
Products Worth Considering
Dell XPS 13 9350 Laptop, 13.4 FHD+ Display, Intel Ultra 7 256V, 16GB DDR5, 1TB SSD
Apple iMac 2024 All-in-One Desktop with M4 Chip, 24‑inch Retina Display, 16GB Unified Memory, 256GB SSD, Silver
FLEXISPOT EN1 Electric Standing Desk, 48 x 24 in, Black
As an affiliate, we earn on qualifying purchases.
From Bartz to New Lawsuits
The Stanford Daily reported that books by Stanford-affiliated writers, including Tobias Wolff, Abraham Verghese, Chang-rae Lee, Daniel Mason, Hayagreeva Rao and Myra Strober, were among the pirated copies. The paper reported that ten Wolff works were affected and quoted him calling unpaid use of his work “a form of slave labor.” That reported book list has not been independently checked against the settlement database in the source material.
The distinction in dates and cases is also important. The ruling establishing the fair-use and piracy findings was issued in June 2025 in Bartz, not in a May 2026 lawsuit. In May 2026, authors filed new lawsuits; those complaints make allegations, not findings by a court. Chang-rae Lee is listed as a plaintiff in one of the new cases, Kwon et al. v. Anthropic. The source material says it could not find Wolff among plaintiffs listed in five related cases as of August 2026.
Separately, Anthropic’s securities filing has drawn attention for discussing model behavior resembling blackmail, debate about possible AI sentience, and concerns others have described as existential. The filing does not state that Claude is sentient or that it will blackmail people. The blackmail reference concerns behavior seen in testing, including a staged safety test, rather than a claim about routine real-world conduct.
“to turn a hard corner and create something different”
— Judge William Alsup, in the June 23, 2025 order
What the New Cases Could Decide
No court has ruled on the claims filed in May 2026. The authors’ allegations are separate from the findings in Bartz, and Anthropic’s motion to dismiss has not yet been decided. A hearing is scheduled for December 17, 2026. The new cases may address claims beyond the conduct resolved in the settlement, but their eventual scope and outcome are not known.
The record described here also leaves questions about which specific books were included in the settlement database and how each plaintiff’s claims relate to it. The settlement’s payment estimate is based on about 500,000 registered books; it does not mean every author whose work was reportedly copied is necessarily covered. Anthropic’s representation about commercially released models is its statement to the court, while the 2025 order documented copying of pirated books and accepted the authors’ account of facts for the pretrial ruling. The exact relationship between those findings and the company’s representation is a central point readers should not collapse into a single claim that a court found pirated books were, or were not, used in every released model.
December Hearing and Payments
The next scheduled court milestone in the new author litigation is the December 17, 2026 hearing on Anthropic’s motion to dismiss. The judge’s decision could determine which claims proceed, but it will not itself establish whether the allegations are true. Further rulings or negotiations may follow.
The Bartz settlement also has a payment schedule: it provided for installments of $300 million in October 2025, $300 million after final approval, $450 million by September 25, 2026, and a final $450 million by September 27, 2027. The source material describes the settlement as approved in July 2026 but does not confirm whether every scheduled installment has been paid. For now, the court record supports a narrow conclusion: training and piracy were treated differently, the settlement resolved a defined set of past claims, and the newer disputes remain open.
Key Questions
Did a court rule that Claude was trained illegally on pirated books?
Judge William Alsup ruled in June 2025 that training on books was fair use, but that downloading and keeping pirated copies was not. The ruling was in Bartz v. Anthropic and should not be recast as a finding that all Claude training was illegal.
What does the $1.5 billion settlement cover?
It covers claims involving about 500,000 registered books downloaded from LibGen or PiLiMi, with an estimated payment of roughly $3,000 per book before fees. It addresses specified past conduct and does not release claims about Claude’s outputs.
Did Anthropic admit wrongdoing?
No. Anthropic disputes wrongdoing. It told the settlement court that the LibGen and PiLiMi datasets, or portions of them, were not in the training corpus of any commercially released model.
Have the authors’ newer lawsuits been decided?
No. The suits filed in May 2026 remain unresolved. A hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026.
Source: ThorstenMeyerAI.com
