🔍 Read the full analysis: Did Anthropic Train Claude On Pirated Books? The Claims And IPO Stakes on ThorstenMeyerAI.com
Get movie nights 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 2025 that Anthropic’s use of books to train Claude was fair use, but that downloading millions of pirated book copies to keep was not. Anthropic later agreed to a $1.5 billion settlement over those downloads while denying wrongdoing; new lawsuits by authors who opted out have not been decided. The company’s IPO filing discusses model risks but does not say Claude is sentient or could blackmail people.
Anthropic’s IPO filing has renewed scrutiny of whether the company used pirated books to train Claude, but the legal record draws a distinction between training and acquiring the books. In June 2025, a federal judge found that training Claude on books was fair use while ruling that Anthropic’s downloads of more than seven million pirated book copies to keep were not; a later $1.5 billion settlement resolved claims by a group of authors, while separate suits by opt-out authors remain pending.
The ruling came in Bartz v. Anthropic, a case filed by three authors in August 2024. In a June 23, 2025 summary-judgment order, Judge William Alsup said using books to train a model was fair use and also found that Anthropic’s purchase and scanning of used print books was fair use. He treated the company’s downloading and permanent retention of pirated copies as a separate issue, finding that conduct unlawful on the facts presented. The case did not go to a jury on damages because the parties later settled.
The court record described 196,640 books from Books3, at least five million copies from Library Genesis and at least two million from the Pirate Library Mirror. The settlement covers rights holders of roughly 500,000 registered books downloaded from LibGen or PiLiMi, at an approximate rate of $3,000 per book before fees. It also requires Anthropic to destroy the downloaded collections and copies, subject to legal-hold obligations. Its release covers past conduct through August 25, 2025, and does not release claims about Claude’s outputs.
Anthropic denies wrongdoing. In settlement materials, it represented that LibGen and PiLiMi datasets, or portions of them, were not in the training corpus of any commercially released large language model. The case record and the company’s statement should not be collapsed into a broader claim that a court found pirated books trained commercially released Claude models: the judge’s decision addressed the downloads and training questions separately, and the settlement did not establish that disputed point through a trial verdict.
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 IPO’s Copyright Questions
The distinction matters to authors, investors and regulators because a finding about how training counts under copyright law is not the same as a finding that acquiring a dataset was lawful. The 2025 order is significant for its treatment of training as fair use on the record before the court, while its ruling on the pirated copies addressed the way Anthropic obtained and retained source material. The settlement put a large dollar figure on one set of claims without settling every dispute about training data or model outputs.
For readers evaluating the IPO coverage, the $1.5 billion settlement is a concrete legal and financial development, but the unresolved opt-out cases could test claims not resolved by that agreement. The company’s S-1 also discusses potential risks from model behavior and public debate over AI, but those disclosures are not findings that Claude is sentient or that it has blackmailed real people. Viral summaries that recast risk language as established conduct go beyond what the filing says.
Popular right now
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Vitamix Propel Series 750 Professional-Grade Blender, 4 Programs, Black
Philips 3300 Series Fully Automatic Espresso Machine with LatteGo Milk System
As an affiliate, we earn on qualifying purchases.
From Bartz to the New Suits
Bartz v. Anthropic began in 2024 and produced the June 2025 order. Judge Alsup said the training use of books was transformative in the case before him, describing it as an effort “to turn a hard corner and create something different.” He also wrote that buying a copy after downloading a pirated one would not erase liability for the earlier theft. Those findings arose at summary judgment, not after a trial on every factual issue.
The parties later agreed to settle the covered claims for $1.5 billion, with payments scheduled in instalments. The settlement was finally approved in July 2026, according to the supplied account. Authors who did not participate in that settlement have filed new cases, including one naming Stanford author Chang-rae Lee, according to the source report. Their claims are distinct from the settled case and have not yet been decided.
A Stanford Daily report named several authors whose books it said were among the pirated copies, including Tobias Wolff, Abraham Verghese, Chang-rae Lee and Daniel Mason. That identification is the student newspaper’s reporting, not a finding established here from the settlement database. The source account also says it could not locate Wolff among plaintiffs listed in five related cases as of August 2026, leaving his status in those actions uncertain.
““to turn a hard corner and create something different””
— Judge William Alsup, in the June 2025 order
Claims Still Before the Courts
The new lawsuits filed by authors who opted out of the settlement have not been decided. A hearing on Anthropic’s motion to dismiss is set for December 17, 2026; the court’s ruling and the scope of any claims that proceed remain unknown. The supplied material does not provide the full complaints or establish what evidence will be considered in those cases.
Some public accounts also combine separate issues: the judge’s 2025 ruling, the settlement’s terms, the newer allegations and Anthropic’s IPO risk disclosures. The reported list of affected Stanford authors has not been independently checked against the settlement database in the source material. Nor does the supplied account establish that every pirated copy identified in the ruling was used to train a commercially released model. Anthropic’s settlement representation about released models and the court record concerning downloaded material must be reported with their different scopes.
The S-1 language about behavior “resembling blackmail” refers to model behavior described as a risk, including a well-known staged safety test, according to the source account. The filing’s mention of debate about AI sentience and concerns raised by others does not amount to a statement that Claude is sentient or that it has blackmailed people. The filing’s full risk disclosures and how investors interpret them remain separate from the copyright cases.
December Hearing and IPO Review
The next listed court milestone is the December 17, 2026 hearing on Anthropic’s motion to dismiss the opt-out authors’ lawsuits. The court may then rule on whether some or all claims can proceed; no outcome is confirmed. The settlement also schedules a final $450 million payment by September 27, 2027, under the terms described in the source account.
As the IPO process continues, investors and authors may focus on how Anthropic describes copyright exposure, training data and model risks in its filings. The filing itself is a disclosure of potential risks, not a court ruling. Further court decisions or more detailed evidence about the new cases could clarify which claims remain live, what datasets are at issue and whether the parties dispute the same conduct covered by the settlement.
Key Questions
Did a judge rule that Anthropic illegally trained Claude on pirated books?
Not in those terms. Judge William Alsup found in June 2025 that training Claude on books was fair use, while ruling that downloading and keeping pirated copies was not. The ruling treated those as separate acts.
How much did Anthropic agree to pay?
Anthropic agreed to a $1.5 billion settlement covering rights holders of roughly 500,000 registered books in the settlement class. The agreement also requires destruction of the downloaded LibGen and PiLiMi books and copies, subject to legal-hold obligations.
Are the newer author lawsuits resolved?
No. The opt-out authors’ cases remain pending, and a hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026. No ruling on those claims is reported here.
Does Anthropic’s IPO filing say Claude is sentient or could blackmail people?
No. According to the source account, the S-1 discusses behavior “resembling blackmail” in a risk context, debate about AI sentience and existential concerns raised by others. Those disclosures do not state that Claude is sentient or that it blackmailed real people.
Source: ThorstenMeyerAI.com
