The Anthropic Copyright Allegations: What’s Known About Claude’s Training
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: The Anthropic Copyright Allegations: What’s Known About Claude’s Training on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

  • Fast, free delivery for dorm and study essentials
  • Prime Video and Amazon Music included
  • Member-only deals
Try Prime for Young Adults Free trial for eligible 18–24 year olds
As an affiliate, we earn on qualifying purchases.

TL;DR

A 2025 court ruling found that training Claude on books and scanning books Anthropic bought were fair use, but that downloading and keeping pirated copies was not. Anthropic later agreed to a $1.5 billion settlement covering about 500,000 books; new lawsuits by authors who opted out remain unresolved.

Anthropic’s copyright dispute has returned to the headlines as authors who did not join its settlement pursue new claims, but the key ruling did not find that training Claude on books violated copyright. In June 2025, U.S. District Judge William Alsup found that using books to train the model was fair use while separately ruling that Anthropic’s downloading and retention of pirated book copies was unlawful; a $1.5 billion settlement later resolved related claims for eligible authors.

In Bartz v. Anthropic, three authors sued in 2024. Alsup’s June 23, 2025 summary-judgment order distinguished among the ways the company obtained books. He found that training on books was fair use and that buying print books and scanning them was fair use. But the order found against Anthropic for obtaining pirated copies and keeping them in a permanent library. The judge described the collections as including 196,640 Books3 titles, at least five million copies from Library Genesis and at least two million from the Pirate Library Mirror.

The ruling was not a final jury verdict on damages. For the piracy claims, Alsup accepted the authors’ account of the facts at the summary-judgment stage, and a jury was expected to address remaining issues. The parties settled before that trial took place. Anthropic agreed to pay $1.5 billion for claims involving roughly 500,000 registered books, or about $3,000 per book before fees. The settlement received final approval in July 2026 and requires the company to destroy the downloaded LibGen and PiLiMi books and copies, subject to legal-hold obligations.

The settlement covers specified past conduct through August 25, 2025, according to the supplied account; it does not release claims over Claude’s outputs. Anthropic denies wrongdoing and represented to the settlement court that neither the LibGen nor PiLiMi datasets, nor portions of them, were in the training corpus of any commercially released model. That statement sits alongside the 2025 ruling’s finding that portions of the pirated collections were used to train some of Anthropic’s models. The distinction between the court’s findings and the company’s settlement representation is central to interpreting the case.

At a glance
reportWhen: The court ruling was issued June 23, 20…
The developmentNew author lawsuits and a public-company filing have renewed scrutiny of Anthropic’s book use, but the central court ruling dates to June 2025 and did not find that Claude’s training was illegal.
Thorsten Meyer AI
Court record

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.

7M+pirated book copies downloaded, per the 2025 court order
$1.5Bclass settlement, final approval July 2026
~$3,000per book, for about 500,000 books
5related opt-out cases now before one judge
  • 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.

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

What the Settlement Covers

The dispute matters to authors and publishers because it sets out how copyright questions can turn on separate steps: acquiring a work, storing it, and using it to train an AI model. Alsup’s ruling did not treat those actions as interchangeable. It found the training use and scans of purchased books fair use, while rejecting the retention of pirated copies. That makes the case more specific than a blanket ruling that AI training is either lawful or unlawful.

The settlement also offers a substantial payment to eligible rights holders, but it is not a general licence for AI companies to use books. Its coverage is limited to the defined claims and books in the case, and authors who opted out can still pursue their own lawsuits. For people using or building AI products, the unresolved claims and the boundaries of the settlement leave questions about how copyright disputes over training data will be handled in future cases.

Anthropic’s filing to go public has added attention to the issue. The supplied source says its S-1 discusses model behavior described as “resembling blackmail”, as well as possible regulation linked to debates over AI sentience and concerns raised by others about existential risks. Those are risk disclosures, not findings that Claude is sentient or that it blackmailed a person. They should not be confused with the separate copyright case.

Sony BRAVIA 5 85-Inch Mini LED 4K HDR Smart Google TV (K-85XR50)

Sony BRAVIA 5 85-Inch Mini LED 4K HDR Smart Google TV (K-85XR50)

Shark PowerDetect Robot Vacuum and Mop Combo with 60-Day Self-Emptying Base (AV2800ZE)

Shark PowerDetect Robot Vacuum and Mop Combo with 60-Day Self-Emptying Base (AV2800ZE)

As an affiliate, we earn on qualifying purchases.

From Bartz to New Lawsuits

The original case, Bartz v. Anthropic, began in August 2024. In June 2025, Alsup drew a distinction between using books to teach a model and keeping pirated copies in a library. The decision addressed fair use, a part of U.S. copyright law that can permit certain uses of copyrighted material without permission. The authors’ remaining claims over the pirated books were settled rather than decided by a jury.

After the settlement, authors who opted out filed new lawsuits in 2026. The source identifies Chang-rae Lee as a named plaintiff in Kwon et al. v. Anthropic, one of the related cases, and says a motion-to-dismiss hearing is scheduled for December 17, 2026. These actions are not the case in which Alsup issued the 2025 ruling. A headline saying a judge ruled in a May 2026 lawsuit that Anthropic illegally trained Claude would conflate the older decision with newer, unresolved claims.

The Stanford Daily reported that books by Stanford-connected writers, including Tobias Wolff, Abraham Verghese, Chang-rae Lee and Daniel Mason, were among the pirated copies. The source material says that list has not been independently checked against the settlement database. It also says Wolff’s name could not be found among the plaintiffs listed in five related cases as of August 2026. Those points remain attributed reporting, not findings established by the court record described here.

““over seven million copies of books””

— Judge William Alsup, describing the pirated book collections

Claims Still Before the Court

The new opt-out lawsuits have not been decided. The court has not ruled on those authors’ claims, and the scheduled December 2026 hearing concerns Anthropic’s motion to dismiss, not a final decision on liability. The outcome could clarify which claims proceed, but it cannot be treated as a ruling that the 2025 case already established.

There is also a factual distinction still requiring careful treatment: the settlement account says Anthropic represented that pirated datasets were not in the training corpus of any commercially released model, while the 2025 order found that portions of the pirated collections were used to train some models. The supplied material does not reconcile the scope or timing of those statements. The Stanford Daily’s reported book list likewise has not been verified here against the settlement database, and the available account does not establish whether Wolff is a plaintiff in a later action or amendment.

December Hearing on Dismissal

The next identified court milestone is the December 17, 2026 hearing on Anthropic’s motion to dismiss the new opt-out claims. The hearing may determine which allegations continue, but the source material does not say when the court will issue a decision or whether the cases will proceed to trial.

Settlement payments are scheduled in instalments, with the final $450 million due by September 27, 2027, according to the supplied account. Readers should distinguish those payment and compliance steps from the separate lawsuits, which concern authors and claims outside the settlement’s release. Any further conclusions about Claude’s training data or the new claims depend on court records and evidence that have not yet produced a ruling.

Key Questions

Did a judge rule that training Claude on books was illegal?

No. In June 2025, Judge William Alsup found that training Claude on books was fair use. He separately ruled against Anthropic for downloading and retaining pirated copies.

What did Anthropic agree to pay?

Anthropic agreed to a $1.5 billion settlement covering claims involving roughly 500,000 registered books. The settlement received final approval in July 2026, according to the supplied source.

Does the settlement license all books for AI training?

No. The settlement addresses specified past claims and books; it is not a general licence to train on copyrighted works. Claims about Claude’s outputs are also not released under the terms described in the source.

What is happening with authors who opted out?

They have filed new lawsuits that remain unresolved. A hearing on Anthropic’s motion to dismiss is scheduled for December 17, 2026; no decision on those claims has been reported in the supplied material.

Source: ThorstenMeyerAI.com

You May Also Like

Column On Parenting: The Challenge Of Doing Everything Vs. Simplifying As A Single Parent

Exploring the challenge parents face between managing every task and simplifying their approach, and why this impacts family well-being.

The cleaner cap table. Why Anthropic’s public-benefit structure dodges OpenAI’s charitable-trust problem — and trades it for a governance question of its own.

Analysis of how Anthropic’s mission-focused governance structure avoids OpenAI’s conversion issues, yet introduces new public market challenges.

Portfolio. The synthesis.

A new synthesis essay by Thorsten Meyer consolidates six institutional approaches to European sovereign AI, guiding policy ahead of August 2026 enforcement.

Grimfaste: Operations for a Fleet

Grimfaste introduces a control platform for managing large publishing fleets, focusing on operational health, link integrity, and privacy compliance in Europe.