Bartz v. Anthropic: The $1.5B AI Copyright Settlement

The Bartz v. Anthropic settlement is a $1.5 billion class action deal — the largest publicly reported copyright recovery in U.S. history — in which Anthropic PBC, the maker of the Claude AI chatbot, agreed to pay roughly $3,000 per book to authors and publishers whose works it downloaded from pirate websites to train its models. The agreement also requires Anthropic to destroy the pirated files. As of mid-2026, it is awaiting final approval in the U.S. District Court for the Northern District of California.¹

What Anthropic Was Accused Of

Authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson filed the case on August 19, 2024, alleging that Anthropic built training datasets for Claude out of copyrighted books it had no right to use.²

Court filings put numbers on the scope. Between 2021 and 2022, Anthropic downloaded more than seven million pirated books from three sources: about 196,640 from a dataset called Books3 in early 2021, at least five million from Library Genesis (LibGen) in June 2021, and at least two million from Pirate Library Mirror (PiLiMi) in July 2022. Internal documents showed employees knew the sources were illegal; one described the choice as a way to avoid the “legal/practice/business slog” of licensing books properly.³

The Ruling That Made the Settlement Happen

Senior U.S. District Judge William Alsup split the legal questions in a June 23, 2025 summary judgment ruling. Using copyrighted books to train Claude was, he found, “exceedingly transformative” and protected as fair use under Section 107 of the Copyright Act.³ Downloading books from pirate sites and keeping them in a permanent library was not. Alsup called that library “inherently, irredeemably infringing.”⁴

In July 2025, Alsup certified a class limited to the piracy claims. The fair use finding on training applied only to the three named plaintiffs, not the class.⁵ With piracy liability on the table and statutory damages potentially reaching $150,000 per willfully infringed work, Anthropic and the plaintiffs signed a term sheet on August 25, 2025.

What the Settlement Pays and Requires

The core financial terms:

A $1.5 billion fund, paid in four installments: $300 million by October 2, 2025; $300 million after final approval; $450 million by September 25, 2026; and $450 million by September 25, 2027. Interest accrues on the later payments.⁶
Roughly $3,000 per qualifying book before deductions for attorneys’ fees and costs. If eligible works exceed 500,000, Anthropic pays an additional $3,000 per additional work.¹
Anthropic must destroy all files it downloaded from LibGen and PiLiMi, plus any derivative copies.⁶

What the deal does not do matters just as much. It releases only past claims tied to the LibGen and PiLiMi downloads. It grants no license for future AI training, releases no claims arising after August 25, 2025, and does not cover claims about the outputs Claude generates.¹ Anthropic did not admit wrongdoing.⁷ The company also represented to the court that the LibGen and PiLiMi datasets were not used in the training corpus of any of its commercially released language models.⁶

Who Qualifies and How Claims Split

The class covers legal or beneficial copyright owners of books Anthropic downloaded from LibGen in June 2021 or PiLiMi in July 2022. Each work must meet three conditions: it has an ISBN or ASIN, it was registered with the U.S. Copyright Office within the required timeframe, and it appears on the court-approved Works List.⁵ International authors, publishers, and estates are eligible if their works are registered with the U.S. Copyright Office.⁸

The final Works List contains 482,460 titles.⁹ For trade and university press titles, the default split is 50/50 between author and publisher. Authors who are sole rights holders — self-published writers or those whose rights have reverted — receive 100% of the award for their work. Claimants can deviate from the default by submitting contract documentation. Educational and textbook titles have no default split and require a good-faith statement of the contractual division.⁵

The claim deadline was March 30, 2026. JND Legal Administration accepted claims online, by email, and by mail.¹⁰ By April 17, 2026, 440,490 of the 482,460 eligible works had been claimed, about 91.3%.⁹

Objections and the Fee Fight

As of the May 2026 hearing, 350 class members had opted out and 53 had filed formal objections.¹¹

Objections clustered around a few points. Some argued $3,000 per book was a “pittance” against potential willful-infringement damages of up to $150,000 per work at trial. Others said the settlement unfairly treats each copyright registration as one work even when a single registration covers multiple books. Objectors also argued that works published under pen names were improperly excluded, disadvantaging small publishers and self-published authors.¹¹ Objector James R. Sills argued the deal fails to restrict future use of the pirated works or verify that destruction is thorough.¹²

Attorneys’ fees drew the sharpest fight. Class counsel initially sought $225 million (15% of the fund), and three additional firms sought another $75 million (5%).¹³ Critics said the combined request could work out to $10,000 to $12,000 per attorney hour.¹² In a December 23, 2025 memorandum, Judge Alsup wrote that the memo “should be read as opposing any award at all for these firms,” questioned whether class counsel had properly allowed “unapproved firms” to do substantial work, and warned that pulling fees from the class fund effectively subsidized certain publishers at other class members’ expense.

Where the Case Stands Now

Judge Alsup retired and took inactive status in late 2025. The case was randomly reassigned to Judge Araceli Martínez-Olguín.¹³

The final approval hearing was held May 14, 2026. Martínez-Olguín did not grant approval that day. She ordered Anthropic to file a brief by May 21 explaining why late opt-outs should not be honored, and told plaintiffs to submit a revised proposed order addressing the objections. She indicated she did not need further input from objectors, and observers reported that final approval could come shortly after the supplemental filings.¹⁴

If final approval holds and no appeal delays distribution, the Authors Guild has estimated payments could begin in late fall 2026 at the earliest. The installment schedule means the last of the money will not arrive from Anthropic until September 2027.⁹

The Opt-Out Path

Some authors decided the class deal was not enough. On December 22, 2025, journalist John Carreyrou and five other writers sued Anthropic along with Google, Meta, OpenAI, xAI, and Perplexity AI in the Northern District of California.¹⁵ Their complaint alleges willful copyright infringement and seeks up to $150,000 per work. The plaintiffs explicitly declined a class structure, saying they did not want to settle “for pennies on the dollar.” The case remains active as of mid-2026.¹⁶

Why This Settlement Matters Beyond the Class

More than 40 copyright lawsuits involving AI companies were pending in U.S. courts as of mid-2026, with defendants including Google, OpenAI, and Meta.¹⁷ Bartz is the only one so far to reach class certification and a settlement.¹⁸ Newer suits are increasingly built around how AI companies acquired their training data rather than whether training itself infringes, and other plaintiffs, including music publishers, have amended complaints to add piracy-based claims.¹⁹

Courts do not agree, though. Two days after Alsup’s ruling, Judge Vince Chhabria held in Kadrey v. Meta Platforms that Meta’s use of copyrighted works to train its Llama model qualified as fair use even where the books came from pirated sources. Chhabria rejected Alsup’s approach of separating the download from the training and said the two should be treated as a single inquiry. Both decisions come from judges in the same district, so neither binds the other, and the split is widely expected to reach the Ninth Circuit.¹⁹

Whether $1.5 billion changes Anthropic’s own trajectory is a separate question. Reports in January 2026 indicated the company was raising a new funding round at a $350 billion valuation.¹⁸

1
Susman Godfrey LLP. Susman Godfrey Secures $1.5 Billion Settlement in Landmark AI Piracy Case
2
CourtListener. Bartz v. Anthropic PBC
3
Publishers Weekly. Federal Judge Rules AI Training Is Fair Use in Anthropic Copyright Case
4
NPR. Anthropic Authors Settlement Pirated Chatbot Training Material
5
Authors Guild. What Authors Need to Know About the Anthropic Settlement
6
ClassAction.org. Bartz et al. v. Anthropic PBC Notice
7
The Lyon Firm. Anthropic AI Copyright Settlement Unauthorized Data Rights
8
Wolters Kluwer. The Bartz v. Anthropic Settlement: Understanding America’s Largest Copyright Settlement
9
Authors Guild. Anthropic Settlement Update: 91 Percent of Books Claimed
10
AnthropicCopyrightSettlement.com. Options and Due Dates
11
Courthouse News Service. Authors, Publishers Near Final Approval of $1.5 Billion Anthropic Copyright Settlement
12
Ars Technica. Authors Fight for Higher Payouts From Anthropic’s $1.5B Copyright Settlement
13
Authors Alliance. Bartz v. Anthropic: Updated Opt-Out and Objection Dates and a New Judge
14
Publishers Weekly. Little Drama at Anthropic’s Settlement Hearing
15
Bloomberg Law. OpenAI, Anthropic, xAI Hit With Copyright Lawsuit From Writers
16
CourtListener. Carreyrou v. Anthropic PBC
17
MIT Technology Review. AI Copyright Meta Anthropic
18
Authors Alliance. AI Class Action Litigation Update: Books, Where Things Stand in Early 2026
19
Norton Rose Fulbright. AI in Litigation Series: An Update on AI Copyright Cases in 2026