🔍 Read the full analysis: Anthropic And Claude: Understanding The Claims About Book Data on ThorstenMeyerAI.com
Get the little things that make your day 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 federal judge ruled in June 2025 that Anthropic’s training on books was fair use, while separately finding that downloading pirated book copies to build a permanent library was not. Anthropic later agreed to a $1.5 billion settlement covering about 500,000 books; it denies wrongdoing and says the LibGen and PiLiMi collections were not used to train commercially released models. New opt-out lawsuits are pending, with a dismissal hearing set for December 17, 2026.
A federal judge ruled in June 2025 that Anthropic’s use of books to train Claude was fair use, while finding that the company’s downloading of more than seven million pirated book copies to create a permanent library was not. The distinction is central to renewed claims about books by Stanford-affiliated writers: a separate set of authors’ lawsuits is pending, and no ruling has been issued on those claims.
The ruling came in Bartz v. Anthropic, a case filed by three authors in August 2024. Judge William Alsup’s June 23, 2025 order separated the act of training a model from how some of the source books were obtained. He found training on books to be fair use and also found that Anthropic’s purchase and scanning of used print books was fair use. But he ruled against the company’s downloading of pirated copies from online collections for permanent storage.
The order described 196,640 books from Books3, at least five million copies from Library Genesis (LibGen), and at least two million from the Pirate Library Mirror (PiLiMi). The decision on piracy was made at summary judgment, before trial. According to the supplied account of the order, the judge accepted the authors’ version of facts for that part of the case; a jury would otherwise have considered remaining issues, including damages. The case settled before that trial took place.
Anthropic agreed to a $1.5 billion settlement, finally approved in July 2026, covering roughly 500,000 registered books downloaded from LibGen or PiLiMi. That works out to about $3,000 per book before fees. The settlement requires destruction of the downloaded books and copies, subject to legal-hold obligations, and releases specified past claims through August 25, 2025. It does not release claims about Claude’s outputs. Anthropic denies wrongdoing and represented to the settlement court that neither dataset, nor any portion of it, was 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
The Difference Between Training and Piracy
The dispute matters because the phrase “illegally trained” can collapse two distinct questions: whether training on copyrighted books qualifies as fair use, and whether a company may copy books from pirated collections to build a library. In the 2025 case, the judge answered those questions differently. The ruling did not find that Claude’s book training was illegal; it found that the permanent acquisition of pirated copies was not excused by fair use.
The settlement also does not amount to a general licence for AI companies to use copyrighted books. It resolves specified claims over past conduct and registered works, while claims about model outputs were expressly left outside the release. For authors, publishers and AI developers, the outcome shows that copyright disputes may turn on the source and handling of training material as well as the use made of it.
For readers evaluating claims about individual writers’ books, the distinction between a court finding and a news report’s account is equally important. The case established facts about specified collections and conduct, but the supplied source says it has not independently checked the Stanford Daily’s book list against the settlement database.
Products Worth Considering
Hyperice Hypervolt 3 Percussion Massage Gun with Heated Head and Premium Carry Case
Oura Ring 5 Smart Ring – Silver – Size 10
Fitbit Versa Smart Watch – Black Aluminum, One Size (S & L Bands Included)
As an affiliate, we earn on qualifying purchases.
How the Separate Cases Relate
The Stanford Daily reported that books by Stanford authors, including Tobias Wolff, Abraham Verghese, Chang-rae Lee, Daniel Mason, Hayagreeva Rao and Myra Strober, were among the pirated copies. It reported that ten of Wolff’s works were affected and quoted him describing unpaid use of his work as “a form of slave labor.” Those details are the student newspaper’s reporting, not findings identified in the 2025 court order.
The supplied source says the Daily’s headline and account risk blending separate timelines. The ruling it cites was issued in June 2025 in Bartz, not in a May 2026 lawsuit. In May 2026, according to the source, authors who opted out of the settlement filed new suits; those claims have not yet been decided. Chang-rae Lee is listed as a plaintiff in one of the new cases, Kwon et al. v. Anthropic. The source says it could not find Wolff among plaintiffs listed in five related cases as of August 2026, while noting he could be included in another action or a later amendment.
The settlement’s scope is also narrower than some descriptions might suggest. It covers qualifying claims tied to past copying through August 25, 2025, and requires destruction of the specified pirated collections. It does not settle every possible copyright dispute involving Claude or authorize future use of books without permission.
““spectacularly so””
— Judge William Alsup, in the June 23, 2025 order
What the New Claims Must Prove
The new lawsuits by authors who opted out of the settlement remain unresolved. The supplied source says a hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026; no decision on those claims had been issued as of October 28, 2026. The lawsuits’ allegations should not be described as findings by a court.
There is also a record distinction around Anthropic’s settlement statement. The source says the 2025 court record found pirated books were copied and that portions of the collections were used to train some Anthropic models, while Anthropic represented that the collections or portions were not in the training corpus of commercially released models. The supplied material does not fully reconcile the scope or timing behind those statements. They concern different descriptions of model training and do not, by themselves, establish that a commercially released Claude model was trained on the collections.
The reported list of Stanford authors’ books has not been independently verified against the settlement database in the supplied account. Wolff’s status as a plaintiff in the related cases is also unresolved there. The precise book-by-book eligibility and the outcome of the opt-out claims remain matters for court records and later proceedings.
December Hearing and Case Updates
The next stated court milestone is the December 17, 2026 hearing on Anthropic’s motion to dismiss the opt-out authors’ claims. The court’s decision on that motion will determine whether those claims proceed, but it will not itself establish that every allegation is true or resolve any damages question.
Readers should watch for court orders, amended plaintiff lists and filings that clarify which books and claims are at issue. The settlement also schedules a final $450 million instalment by September 27, 2027, under the payment timetable described in the source. Until further rulings or verified records clarify the claims, the 2025 fair-use decision, the separate piracy finding, Anthropic’s settlement representations and the new allegations should be treated as distinct parts of the story.
Key Questions
Did a judge rule that Anthropic illegally trained Claude on books?
No. In June 2025, Judge William Alsup ruled that training Claude on books was fair use. He separately found that downloading pirated copies to build a permanent library was not fair use.
What does the $1.5 billion settlement cover?
It covers qualifying claims involving roughly 500,000 registered books downloaded from LibGen or PiLiMi. The settlement requires destruction of the downloaded collections and releases specified past claims, but it does not release claims about Claude’s outputs.
Does the settlement prove that Claude’s commercial models trained on pirated books?
Not by itself. Anthropic represented to the settlement court that the LibGen and PiLiMi datasets, or portions of them, were not in the training corpus of commercially released models. The supplied source also describes court findings about copying and training-related conduct, but does not fully reconcile the scope of those statements.
Have the opt-out authors won their new lawsuits?
No ruling on the new claims is reported. A hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026, according to the supplied source.
Source: ThorstenMeyerAI.com
