🔍 Read the full analysis: Did Pirated Books Train Claude? The Allegations Around Anthropic on ThorstenMeyerAI.com
Get hardware and tech essentials 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 found Anthropic’s training on books to be fair use, but held that downloading and keeping pirated book collections was not. Anthropic later agreed to a $1.5 billion settlement covering about 500,000 books; new lawsuits by authors who opted out remain pending, with a dismissal hearing set for December 17, 2026.
Anthropic’s use of pirated books is back under scrutiny as new lawsuits by authors who opted out of a class settlement proceed, but no court has ruled on those newer claims. A June 2025 decision found that using books to train Claude was fair use, while separately ruling that downloading and keeping pirated copies was not; the company later agreed to a $1.5 billion settlement covering about 500,000 books.
The distinction at the center of the dispute is between training on books and acquiring and retaining unauthorized copies. In Bartz v. Anthropic, U.S. District Judge William Alsup ruled on June 23, 2025, that training Claude on books was fair use. He also found that scanning books Anthropic had purchased was fair use. But he rejected the company’s use of pirated collections to build a permanent digital library.
The court order described 196,640 books from Books3, at least five million copies from Library Genesis and at least two million from Pirate Library Mirror: more than seven million copies in total. The judge said that buying a copy later would not erase liability for the earlier taking. The ruling was made before trial and accepted the authors’ account of the relevant facts about the pirated files; the case settled before a jury decided damages.
Under the settlement, finally approved in July 2026, Anthropic agreed to pay $1.5 billion for claims involving roughly 500,000 registered books downloaded from LibGen or PiLiMi, or about $3,000 per book before fees. The agreement requires the company to destroy those downloaded books and copies, subject to legal-hold obligations. It releases specified claims about past conduct through August 25, 2025, but does not release claims concerning Claude’s outputs. Anthropic denies wrongdoing and said in settlement filings that the pirated collections were not 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 Line Between Training and Piracy
The case matters because the court did not treat every use of a copyrighted book alike. It found that training could qualify as fair use while holding that acquiring and keeping unauthorized copies was not. That distinction is central to copyright disputes over generative AI: a court’s view of model training does not automatically settle questions about how training material was obtained or stored.
The settlement also gives authors a substantial recovery for the covered books, but it is not a general license for AI companies to use copyrighted material. Its scope is limited to particular past claims and registered works. Anthropic’s statement that the pirated datasets were not in commercially released models’ training corpora is a company representation in settlement filings; it does not erase the court’s findings about downloads or resolve the separate claims by authors who opted out.
The dispute has practical stakes for writers, publishers and AI developers. Authors want compensation and limits on unlicensed copying; developers face legal uncertainty over datasets and training practices. For readers, the key point is that the 2025 ruling did not declare all AI training lawful, nor did it find that Claude was trained on every book in the pirated collections.
Products Worth Considering
LG UltraGear 27G640A-B 27-Inch QHD IPS Gaming Monitor, 300Hz, 1ms, G-SYNC Compatible, FreeSync Premium, DisplayHDR 400
Apple MacBook Pro 16.2-inch with M5 Pro, 18-core CPU, 20-core GPU, 48GB Memory, 1TB SSD, Wi-Fi 7, Space Black
Trezor Safe 7 Crypto Hardware Wallet
As an affiliate, we earn on qualifying purchases.
What the 2025 Court Decided
Bartz v. Anthropic was brought by three authors in August 2024. Judge Alsup’s June 2025 order drew separate conclusions about books Anthropic purchased and scanned, books used for training, and pirated copies downloaded from online collections. He described the model training as transformative and found it fair use, while finding the retention of pirated copies outside that protection.
The piracy question did not reach a jury trial. Anthropic and the authors reached the $1.5 billion settlement, which was finally approved in July 2026. The settlement concerns a defined set of books and past conduct; it does not settle all disputes about training data or model outputs.
Authors who opted out filed new lawsuits in May 2026. A report by The Stanford Daily identified Stanford-affiliated writers whose works it said appeared among the pirated copies. The supplied account says the book list has not been independently checked against the settlement database, and the plaintiff lists reviewed through August 2026 did not include Tobias Wolff in the five related cases. Chang-rae Lee was identified as a named plaintiff in one of the new actions. Those details should be treated as reported, not as findings by a court.
“The company represented that 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, in settlement filings
What the New Claims Allege
The opt-out lawsuits have not been decided, and a May 2026 filing is not itself a court ruling. A hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026. The court record described in the source material does not establish how that motion will be decided or whether the cases will proceed to trial.
There is also a difference between evidence that pirated collections were downloaded and the company’s statement that those collections were absent from commercially released models’ training corpora. The 2025 case settled before a jury resolved the remaining issues. The settlement’s terms, the judge’s findings, and Anthropic’s representation address different questions and should not be collapsed into a single claim that Claude was trained on every pirated book, or that it was trained on none.
Reports naming individual authors and books also need care: the supplied source says those lists were not checked against the settlement database, and the status of some authors in the related cases was uncertain as of August 2026. Separately, viral descriptions of Anthropic’s public filing as saying Claude could become sentient or end humanity go beyond the account of the filing provided here. The source says the filing discusses behavior resembling blackmail in a staged safety test, debate about possible AI sentience and concerns raised by others about existential risks.
December Hearing and Case Status
The next scheduled court milestone is the December 17, 2026 hearing on Anthropic’s motion to dismiss the opt-out authors’ lawsuit. A hearing date does not determine the result; the judge’s decision and any further case schedule remain pending.
The settlement also includes scheduled payments, with the final $450 million installment due by September 27, 2027, according to the supplied account. The parties and court may provide further updates about administration of claims, book registration and the new cases. Until then, readers should distinguish the settled claims about past downloads from the unresolved allegations and avoid treating either side’s account of training data as a new court finding.
Key Questions
Did a court find that Claude was trained illegally on pirated books?
The June 2025 court ruling found that training on books was fair use, but that downloading and keeping pirated copies was not. It did not decide that every book in those collections was used to train a commercially released version of Claude.
What does Anthropic’s $1.5 billion settlement cover?
The settlement covers specified past copyright claims involving roughly 500,000 registered books downloaded from LibGen or PiLiMi. It requires destruction of the downloaded copies, subject to legal-hold obligations, and does not release claims about Claude’s outputs.
Have the authors who opted out won their new lawsuits?
No. The new lawsuits remain pending, and no ruling on their claims is described in the supplied material. A hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026.
Did Anthropic’s public filing say Claude is sentient or could end humanity?
Not according to the source account. It says Anthropic’s filing discusses model behavior resembling blackmail in a staged test, debate about possible AI sentience and existential concerns raised by others. Those are not the same as a statement that Claude is sentient or will end humanity.
Source: ThorstenMeyerAI.com
