🔍 Read the full analysis: A Reader’s Guide To The Anthropic Pirated Books Claims on ThorstenMeyerAI.com
Get the latest gadgets 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 distinguished Anthropic’s use of books to train Claude from its downloading and retention of millions of pirated copies. Anthropic later agreed to a $1.5 billion settlement covering roughly 500,000 books, while new opt-out authors’ lawsuits remain pending. Anthropic denies wrongdoing and says the pirated collections were not used to train any commercially released model.
The record behind claims that Anthropic trained Claude on pirated books is more specific than some headlines suggest: a federal judge ruled in June 2025 that training on books was fair use, but that downloading and keeping millions of pirated copies was not. Anthropic later agreed to a $1.5 billion settlement covering about 500,000 books, while separate lawsuits brought by authors who opted out of that settlement have not yet been decided.
In Bartz v. Anthropic, U.S. District Judge William Alsup’s June 23, 2025, order drew a distinction between how books were used and how Anthropic obtained some of them. He said training Claude on books was fair use, describing the use 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.
The ruling treated the company’s downloaded collections differently. It described 196,640 books from Books3, at least five million copies from Library Genesis, and at least two million from the Pirate Library Mirror: more than seven million copies in total. Alsup found that keeping those pirated copies in a permanent library was not fair use. The ruling was a pretrial decision, not a jury verdict; the authors’ account of the relevant facts was accepted for that ruling, and the case settled before a jury decided damages.
The settlement, finally approved in July 2026, provides $1.5 billion for rights holders associated with roughly 500,000 registered books. The source account describes payments in four instalments and says Anthropic must destroy the downloaded LibGen and PiLiMi books and copies, subject to legal-hold obligations. Anthropic denies wrongdoing and represented to the settlement court that neither collection, nor portions of them, appeared in the training corpus of any commercially released large language 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 distinction matters because copyright liability depends on the conduct at issue. The 2025 order did not rule that training Claude on books was unlawful; it found that training use to be fair use on the record before it. It did find that building and retaining a library from pirated downloads was not fair use. Treating those findings as one blanket ruling that “Anthropic illegally trained Claude” can misstate what the judge decided.
The settlement also does not grant Anthropic permission to train on books generally. According to the source account, it resolves specified past claims through August 25, 2025, while claims about Claude’s outputs are not released. For authors and publishers, the case puts a large monetary value on the dispute over pirated copies without settling every question about AI training, licensing or generated material.
Popular right now
Roborock Q7 M5+ Robot Vacuum and Mop
Philips 5500 Series Fully Automatic Espresso Machine with LatteGo
Dyson Cyclone V10 Origin Cordless Vacuum
As an affiliate, we earn on qualifying purchases.
How the Cases Fit Together
The original Bartz case was filed by three authors in August 2024. It produced Alsup’s June 2025 order and then the settlement. The settlement’s scale—about $3,000 per registered book before fees, based on $1.5 billion divided across roughly 500,000 books—is a useful description of its aggregate terms, not a guaranteed individual payment amount.
A separate group of authors who opted out filed new cases in May 2026, according to the source material. Those filings are allegations, not findings by a judge. The source reports that Chang-rae Lee is named as a plaintiff in one action, Kwon et al. v. Anthropic. It also says The Stanford Daily reported that books by several Stanford-affiliated authors appeared among the pirated copies, including works by Tobias Wolff, Abraham Verghese, Daniel Mason, Hayagreeva Rao and Myra Strober. The book list has not been independently checked against the settlement database in the source account.
The original article also addressed social-media claims about Anthropic’s public offering filing. Its account says the S-1 does not state that Claude is sentient or that it could blackmail users. It says the filing discusses model behavior “resembling blackmail,” with a prominent example arising in a staged safety test, and refers to debate over possible AI sentience and concerns others have raised about existential risk.
““neither the LibGen or PiLiMi datasets, nor any portions of those datasets, were in the training corpus of any of its commercially released large language models.””
— Anthropic, as described in its representation to the settlement court
What the New Suits Could Establish
No ruling has been issued on the opt-out authors’ May 2026 claims, and the source account says a hearing on Anthropic’s motion to dismiss is set for December 17, 2026. The claims could be narrowed or dismissed, or proceed; the eventual outcome is not known.
There is also a point of tension between the 2025 court record and Anthropic’s later settlement representation. The source says the court record showed pirated books were copied and portions of the collections were used to train some Anthropic models, while Anthropic represented that the collections and their portions were not in the training corpus of any commercially released model. Those statements concern different scopes—some models versus commercially released models—and should not be treated as identical claims.
The Stanford Daily’s reported list of affected books and authors has not been independently verified against the settlement database in the source account. The account also says it could not find Wolff among plaintiffs listed in five related cases as of August 2026; his involvement in another action or a later amendment remains possible. No court has ruled that the newly named authors’ specific claims are true.
The December Hearing and Claims
The next identified court milestone is the December 17, 2026, hearing on Anthropic’s motion to dismiss the opt-out authors’ case. Court action on that motion should clarify which claims, if any, move forward; it will not by itself decide all factual questions or establish damages.
The settlement’s remaining payment and destruction obligations also remain relevant to how the original case is completed. Readers should distinguish any future court findings from the parties’ arguments and from reported book lists. Until the new lawsuits are decided, the confirmed record remains the 2025 ruling’s split treatment of training and pirated-copy retention, alongside a settlement that resolved specified past claims without deciding every dispute over AI-generated outputs.
Key Questions
Did a judge rule that Anthropic illegally trained Claude on books?
Not in those terms. In June 2025, Judge William Alsup ruled that training Claude on books was fair use, while finding that downloading and retaining pirated book copies was not. The order was not a jury verdict on damages.
What does the $1.5 billion settlement cover?
The settlement covers claims tied to roughly 500,000 registered books downloaded from LibGen or PiLiMi, according to the source account. It resolves specified past claims; it is not a general licence to train on books, and claims about Claude’s outputs are not released under the terms described.
Has a court ruled on the authors who opted out?
No. The new lawsuits filed in May 2026 remain unresolved. The source account says a hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026.
Does Anthropic say pirated books trained a released Claude model?
Anthropic represented to the settlement court that neither the LibGen nor PiLiMi collections, including portions of them, were in the training corpus of any commercially released model. That is the company’s position; it should be distinguished from the court’s findings about downloads and from claims about other models or stages of development.
Source: ThorstenMeyerAI.com
