🔍 Read the full analysis: What’s The Evidence On Anthropic’s Alleged Use Of Pirated Books? on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A federal judge’s 2025 ruling distinguished Anthropic’s use of books to train Claude, which the judge found was fair use, from the company’s downloading and retention of millions of pirated book copies, which the ruling found unlawful. Anthropic later agreed to a $1.5 billion settlement covering about 500,000 books, while new lawsuits filed by authors who opted out have not been decided.
A federal judge ruled in June 2025 that Anthropic’s use of books to train Claude was fair use, but that downloading and keeping millions of pirated book copies was not. The distinction is central to current claims about the AI company: a later $1.5 billion settlement addressed past copying claims involving about 500,000 books, while separate lawsuits by authors who opted out remain unresolved.
The ruling came in Bartz v. Anthropic, a case brought by three authors in 2024. In an order dated June 23, 2025, U.S. District Judge William Alsup said training Claude on books was fair use. He also found that Anthropic’s purchase of used print books, followed by scanning them, was fair use. But he treated the company’s acquisition and permanent storage of pirated copies as a separate act: the order described more than seven million copies from the Books3, Library Genesis and Pirate Library Mirror collections, and found that the copying was not excused by fair use.
The ruling was a pretrial decision, not a jury verdict on damages. On the pirated-copy issue, Alsup accepted the authors’ version of the facts for the ruling; a jury was expected to address remaining questions, including damages. The case settled before that trial. The settlement, finally approved in July 2026, sets aside $1.5 billion for owners of roughly 500,000 registered books that Anthropic downloaded from LibGen or PiLiMi. The reported payment works out to about $3,000 per book before fees. The agreement also requires destruction of the downloaded collections and copies, subject to legal-hold obligations.
Anthropic has denied wrongdoing. In settlement filings, it represented that neither the LibGen nor PiLiMi datasets, nor portions of them, were in the training corpus of any commercially released large language model. That statement is part of the settlement record; it does not erase the judge’s finding about the downloads, nor does the settlement amount amount to a jury’s damages verdict. The agreement releases specified claims over past conduct through August 25, 2025, but does not release claims about Claude’s outputs.
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 Divide Between Training and Piracy
The case matters because “training on copyrighted books” and “obtaining pirated copies” are legally distinct issues in the ruling. Alsup found the former fair use on the record before him, while finding the latter unlawful. Collapsing those conclusions into a claim that a court found all AI training illegal misstates the decision; saying the court cleared every part of Anthropic’s conduct would also be inaccurate.
The settlement sets a substantial financial and practical consequence for the alleged copying, including payments to eligible book rights holders and destruction of the downloaded collections. At the same time, it does not establish that Claude’s commercial models were trained on those specific collections: Anthropic made the opposite representation in the settlement. The distinction affects authors’ ability to pursue claims, companies’ handling of training data, and public understanding of what copyright rulings have—and have not—decided about generative AI.
Products Worth Considering
DJI Osmo Pocket 3 Filter Bundle – Vlogging Camera with 1 CMOS Sensor and 4K/120fps
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Jackery Solar Generator 1000 v2 with 200W Solar Panel
As an affiliate, we earn on qualifying purchases.
How the Cases Reached This Point
Bartz v. Anthropic began in August 2024, when three authors sued over the company’s use of books. The June 2025 order addressed two different routes by which books entered Anthropic’s process: purchased print books that were scanned, and pirated digital copies kept in a library. Judge Alsup found the scanning and model training at issue to be fair use, but not the downloading and retention of pirated copies.
The settlement resolved the covered claims without a trial on damages. Authors who opted out later brought separate lawsuits. A report by The Stanford Daily named Stanford-affiliated writers whose books it said were included among pirated copies, including Tobias Wolff, Abraham Verghese, Chang-rae Lee and Daniel Mason. Those names and the reported book list should be treated as the newspaper’s reporting unless confirmed against the relevant court or settlement records. The source account says Chang-rae Lee is named in one new case; it could not locate Wolff among plaintiffs in five related cases listed as of August 2026.
The same source account notes that Anthropic filed to go public one day before the newspaper’s report. Viral claims about its filing also circulated, but the account’s review of the S-1 says it does not state that Claude could become sentient or blackmail people. It describes model behavior resembling blackmail in a staged safety test, discusses debate about AI sentience as a possible regulatory issue, and refers to existential concerns raised by others.
““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 represented in settlement filings
What the New Suits Could Decide
The new cases filed by authors who opted out have not been decided. The source account says a hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026; the claims, defenses and any ruling remain pending. The earlier settlement does not resolve claims about Claude’s outputs, and its release applies only to specified past conduct through August 25, 2025.
There are also limits to what can be concluded from the public record described here. The settlement covers registered books and the reported book list in the newspaper account has not been checked against the settlement database. Anthropic’s statement about the training corpora of commercially released models is its own representation, while the 2025 ruling addressed copying and training questions on the record in that case. The source material does not establish how every downloaded book was used across every model or development stage.
The December Hearing and Payments
The next listed court milestone for the opt-out authors’ cases is the December 17, 2026 hearing on Anthropic’s motion to dismiss. The court’s handling of that motion will determine whether some or all claims proceed; no outcome is established yet.
Settlement payments are scheduled in installments, including $450 million by September 27, 2027, according to the settlement terms summarized in the source account. Eligible book owners’ claims and payments remain governed by the settlement process. Further reporting should distinguish those payments and the required destruction of the downloaded collections from the unresolved lawsuits and from any separate disputes over AI-generated outputs.
Key Questions
Did a judge rule that Anthropic illegally trained Claude on books?
No. In June 2025, Judge William Alsup found that training Claude on books was fair use on the record before him. He separately found that downloading and retaining pirated copies was not protected by fair use.
What did the $1.5 billion settlement cover?
The settlement covers specified past claims involving about 500,000 registered books downloaded from LibGen or PiLiMi. It includes payments to eligible rights holders and requires destruction of the downloaded collections and copies, subject to legal-hold obligations.
Did Anthropic say pirated collections trained commercial Claude models?
No. In settlement filings, Anthropic represented that neither the LibGen nor PiLiMi datasets, or portions of them, were in the training corpus of any commercially released large language model. That is the company’s representation in the settlement record.
Are the lawsuits filed by authors who opted out settled?
They remain pending, according to the source account. A hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026; the court has not yet ruled on those claims.
Does the settlement cover claims about Claude’s outputs?
No. The settlement’s release covers specified past conduct through August 25, 2025, and the source account says claims about Claude’s outputs are explicitly not released.
Source: ThorstenMeyerAI.com
