🔍 Read the full analysis: Why Anthropic’s Alleged Use Of Pirated Books Is Drawing Attention on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A 2025 federal court ruling found that Anthropic’s downloading of millions of pirated books to build a permanent library was not fair use, while ruling that training Claude on books was fair use. A $1.5 billion settlement resolved claims by a group of authors, but new lawsuits filed by authors who opted out have not been decided.
Anthropic’s use of pirated books is drawing renewed scrutiny after new author lawsuits followed a 2025 court ruling and a $1.5 billion settlement over copied books. The ruling found that training Claude on books was fair use, but that downloading more than seven million pirated book copies to build a permanent library was not; the new claims have not yet been decided.
The distinction is central to the dispute. In Bartz v. Anthropic, U.S. District Judge William Alsup ruled on June 23, 2025, that Anthropic’s use of books to train its models qualified as fair use. He separately found that acquiring pirated digital copies from online collections and keeping them in a permanent library was not fair use. The ruling was a pretrial decision, not a jury verdict on all issues.
The court’s order described 196,640 books from Books3, at least five million copies from Library Genesis, or LibGen, and at least two million from Pirate Library Mirror, or PiLiMi. The piracy-related claims did not proceed to a jury trial because Anthropic later settled. The source account says Anthropic does not deny the downloads, but the question of damages was resolved through the settlement rather than a trial verdict.
The settlement, approved in July 2026, provides $1.5 billion for owners of roughly 500,000 registered books downloaded from LibGen or PiLiMi—about $3,000 per book before fees. It also requires Anthropic to destroy the downloaded books and copies, subject to legal-hold obligations. The settlement covers specified past conduct, not claims about Claude’s outputs or a general licence for future training.
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
Why the Training Distinction Matters
The case has become a test of how copyright law treats different stages of AI development. The court distinguished learning from books to train a model from obtaining and retaining pirated copies. That means the ruling does not establish that every use of copyrighted material to train AI is unlawful, nor does its fair-use finding make the acquisition of pirated source copies lawful.
The distinction matters to authors, publishers and AI companies because it separates two questions often compressed into the phrase “trained on pirated books”: how a company obtained its source material, and whether using that material to train a model is permitted. The settlement offers a large payment to eligible book rights holders, but it does not settle every author’s claim. Authors who opted out are pursuing separate cases, leaving further court decisions possible.
There is also a business and disclosure angle. The story is receiving attention as Anthropic prepares to go public, but the company’s filing should not be read as a court finding about the book dispute. According to the source account, its S-1 discusses behavior resembling blackmail in a test and concerns raised by others about AI risks; it does not say Claude is sentient or that it will blackmail people. Those distinctions matter when readers evaluate claims based on legal documents or public-company disclosures.
Popular right now
De'Longhi Magnifica Evo Fully Automatic Espresso Machine with Manual Milk Frother
iRobot Roomba Max 715 Pet Robot Vacuum with AutoEmpty Dock
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
As an affiliate, we earn on qualifying purchases.
From Bartz Ruling to New Suits
Bartz v. Anthropic began in August 2024, when three authors sued over the company’s use of their works. In its June 2025 order, the court treated the training and the creation of a permanent digital library as separate acts. It found that scanning print books Anthropic had bought was fair use, as was training the model on books; it rejected the argument that downloading pirated copies for permanent retention was excused by fair use.
The court’s ruling on the pirated collections came at the summary-judgment stage and accepted the authors’ account of the relevant facts. A jury was expected to address remaining matters, including damages, but the settlement ended that path. Under the reported settlement terms, Anthropic said that neither the LibGen and PiLiMi datasets nor portions of them were in the training corpus of any commercially released model. The source account also says the court record described portions of the collections being used to train some models. The settlement did not resolve that tension through a trial finding.
In May 2026, authors who did not join the settlement filed new lawsuits. That filing date is not the date of the 2025 ruling. One of the new cases named in the source account is Kwon et al. v. Anthropic, which lists author Chang-rae Lee as a plaintiff. The Stanford Daily separately reported that books by several Stanford-affiliated writers were among the copies at issue; the source account says it had not independently checked that list against the settlement database.
““spectacularly so””
— Judge William Alsup, in the June 23, 2025 order
Claims Still Before the Courts
The new lawsuits remain unresolved. The source account says a hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026; no ruling on the new claims had been issued as of October 28. The claims in those cases should not be described as findings that Anthropic has already been held liable for the conduct alleged in them.
Some details also remain contested or unverified in the available account. Anthropic disputes wrongdoing and made a representation about the pirated collections not appearing in commercially released models’ training corpora. The earlier court record, as described by the source, addresses use of portions of the collections in training some models. The precise relationship between those records and the settlement representation has not been settled by a trial verdict. The reported list of affected Stanford authors and books has not been independently checked against the settlement database by the source author.
The settlement’s payment schedule includes instalments due through September 2027, so the full payment process is not yet complete. The source material also does not establish the ultimate outcome of the opt-out cases or whether the court will allow them to proceed.
December Hearing and Settlement Payments
The next stated court milestone is the December 17, 2026 hearing on Anthropic’s motion to dismiss the opt-out authors’ claims. The judge’s decision could determine which parts of those cases continue, but it will not by itself establish liability on any claim that survives.
Separately, settlement payments are scheduled in instalments, including a reported $450 million payment by September 27, 2027. The remaining public record to watch includes the court’s handling of the new cases, any updates to the settlement administration, and further filings that clarify the authors and books covered. Until then, the established distinction remains: the 2025 court found training fair use but rejected the permanent retention of pirated copies as fair use.
Key Questions
Did a judge rule that Anthropic’s Claude training was illegal?
No. In June 2025, Judge William Alsup found that training on the books at issue was fair use. He separately ruled that downloading pirated copies to build a permanent library was not fair use.
What does the $1.5 billion settlement cover?
The reported settlement pays owners of roughly 500,000 registered books that Anthropic downloaded from LibGen or PiLiMi. It covers specified past conduct and does not grant a general licence for future training or release claims about Claude’s outputs.
Have the authors’ newer lawsuits been decided?
No. The source account says the suits filed in May 2026 remain unresolved and that a hearing on Anthropic’s motion to dismiss is set for December 17, 2026.
Did Anthropic say pirated books were used in a released Claude model?
Anthropic represented in the settlement record that the LibGen and PiLiMi datasets, or portions of them, were not in the training corpus of commercially released models. The source account says the 2025 court record described portions of those collections being used to train some models. The settlement did not resolve that issue through a trial verdict.
Source: ThorstenMeyerAI.com
