Anthropic’s $1.5B Author Payout: Eligibility, Amounts, and Approval

Anthropic’s $1.5 billion author settlement resolves a class-action copyright lawsuit brought by writers whose books were downloaded from pirate websites and fed into the training data for Claude, the company’s AI chatbot. Under the deal, Anthropic will pay at least $1.5 billion plus interest into a fund covering roughly 482,000 eligible books, which works out to an estimated $2,932 to $3,100 per work.¹ It is the largest copyright settlement in American history. As of mid-2026, the agreement has preliminary approval and is awaiting a final order, with payments expected no earlier than late fall 2026.

What Anthropic Is Actually Paying For

The settlement is not a payment for AI training. That distinction matters, because it defines who gets paid and why.

In June 2025, Senior U.S. District Judge William Alsup issued the first American ruling directly on whether training a generative AI model on copyrighted books qualifies as fair use.² He split the case in two. Training Claude on lawfully acquired books, he held, was “exceedingly transformative” and protected by fair use.³ Downloading more than 7 million books from pirate repositories — mostly Library Genesis (LibGen) and the Pirate Library Mirror (PiLiMi) — and keeping them in a permanent internal library was not.⁴ “There is no carveout from the Copyright Act for AI companies,” the court wrote.⁵

The $1.5 billion pays for the piracy, not the training. Under the Copyright Act, willful infringement can carry statutory damages of up to $150,000 per work, so theoretical exposure across hundreds of thousands of pirated titles reached tens of billions of dollars.⁶

Who Qualifies for a Payment

The settlement class covers legal or beneficial copyright owners of books Anthropic downloaded from LibGen or PiLiMi. To be eligible, a book must meet three conditions:

It must appear on the official settlement Works List.
It must have an ISBN or ASIN.
It must have been registered with the U.S. Copyright Office within five years of publication — either before Anthropic’s download or within three months of first publication.⁷

Authors, publishers, estates, co-authors, and international rightsholders are all eligible regardless of nationality, so long as the book meets those requirements.⁸

The gap between what Anthropic pirated and what the settlement covers is large. Anthropic downloaded roughly 5 million books from LibGen, 2 million from PiLiMi, and nearly 200,000 from the Books3 dataset.⁴ The class comes out to only about 482,000 works. Many pirated files lacked ISBNs, and metadata analysis showed that less than half of the fiction files in the LibGen dataset even contained identifiers.⁹ European writers and publishers whose books were never registered in the United States are excluded entirely.

Participation from those who do qualify has been high. As of May 2026, about 93% of eligible works — roughly 448,000 of 482,000 — had been claimed. Only 350 class members opted out, and 53 filed formal objections.¹⁰

How Much Each Book Is Worth

After deductions, the net fund available to class members is roughly $1.29 billion. Divided across the eligible works, that produces an estimated $2,932 to $3,100 per book, likely rising once interest is added.¹

The deductions from the $1.5 billion gross include:

Attorneys’ fees of 12.5% — about $187.5 million — well below the 25% to 33% typical in class actions of this size.
Litigation expenses of roughly $2.78 million.
An $18.2 million cost reserve for claims administration, a Special Master, and an industry working group.
Service awards of $50,000 each to the three named plaintiffs, Andrea Bartz, Kirk Wallace Johnson, and Charles Graeber.¹

How the per-book payment gets split depends on the book. For trade and university press titles, the default is 50/50 between author and publisher. Co-authors divide the author’s share equally. Self-published authors, and authors whose rights have reverted, receive 100% of the per-work payment. Educational works follow contract-driven splits. If co-owners disagree, a court-appointed Special Master resolves the dispute.¹¹

Payment Schedule and When Money Arrives

Anthropic is paying in four installments:

$300 million deposited in October 2025 (already in an interest-bearing escrow account).
$300 million due five days after final approval.
$450 million on the first anniversary of preliminary approval (September 2026).
$450 million on the second anniversary (September 2027).¹¹

Class members should not expect checks immediately after final approval. Payments to rightsholders are expected to begin no earlier than late fall 2026, after any appeals are resolved.¹

What Anthropic Must Do Beyond Paying

The deal is not only about money. Anthropic must destroy all datasets containing the pirated works, including the original files torrented from LibGen and PiLiMi and any copies derived from them.¹²

The release Anthropic buys is narrow. It covers only past claims tied to the downloading and storage of these specific pirated books. It grants Anthropic no license for future AI training on those works, releases no claims arising after August 25, 2025, and releases no claims based on Claude’s outputs.¹²

Where Approval Stands Right Now

The path to approval has not been smooth. When the parties first brought the deal to Judge Alsup on September 8, 2025, he refused to grant preliminary approval, calling the proposal “nowhere close to complete” and saying he felt “misled.” The parties had no definitive list of covered works, no finalized notice process, and no completed claim form. Alsup also objected to an “army” of add-on attorneys from author and publisher organizations, and ordered that they would not be paid from settlement funds.¹³

He also pressed on value. With statutory damages potentially reaching $150,000 per work, Alsup noted that in many class actions members “get the shaft,” and pushed counsel to show that around $3,000 per book was fair.⁶

The parties revised and resubmitted with a finalized Works List, and Alsup granted preliminary approval on September 25, 2025.¹⁰ The case was later reassigned to Judge Araceli Martínez-Olguín, who held a 75-minute fairness hearing on May 14, 2026, in San Francisco. She questioned counsel on attorneys’ fees and the cost reserve and ordered Anthropic to explain why late opt-outs should not be honored.¹⁴

Objectors have raised several issues: that the settlement counts group copyright registrations as a single work rather than counting each book individually, that works published under pseudonyms are excluded, that a one-time payment is inadequate because Anthropic continues to profit from models trained on the pirated material, and, in one filing, that key settlement documents were uploaded late and the opt-out period should be reopened.¹⁰

As of mid-June 2026, Judge Martínez-Olguín has not issued a final approval order, though reporting suggests one is expected soon.¹⁵

What This Settlement Does Not Cover

Several things the settlement might seem to address, it does not.

It does not decide whether AI training on copyrighted books is legal. Judge Alsup found that training on lawfully acquired books is fair use; the settlement pays only for the pirated acquisition and storage.⁵

It does not license Anthropic to keep using those books, and it does not release claims tied to what Claude generates as output.¹²

It does not reach unregistered or non-ISBN works, and it does not cover authors and publishers outside the U.S. registration system.⁷

And it does not extend to other AI companies. Separate cases against Meta, Microsoft and OpenAI, and Anthropic itself (a still-active suit from music publishers over Claude’s reproduction of song lyrics) are proceeding on their own tracks with different facts and different reasoning.¹⁶

If your book meets the class definition, the claim you file here settles only what Anthropic did with those pirated copies through August 25, 2025. Anything Anthropic does with your work after that date, and anything any other AI company has done with it, remains a separate legal question.

1
Authors Guild. Anthropic Settlement Update: 91 Percent of Books Claimed
2
The Guardian. Anthropic Did Not Breach Copyright When Training AI on Books Without Permission, Court Rules
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
Justia. Bartz et al v. Anthropic PBC, Order on Summary Judgment
6
HBSR. What AI Companies Can Learn From Anthropic’s $1.5 Billion Proposed Class Action Settlement
7
Copyright Alliance. Participating in the Bartz v. Anthropic Settlement
8
AAL/IT Agents. Anthropic Settlement Claims Information for Agents and Creators
9
Authors Alliance. Bartz v. Anthropic: A Preliminary Look at What LibGen Books May Be Included in the Class Action
10
Courthouse News. Authors, Publishers Near Final Approval of $1.5 Billion Anthropic Copyright Settlement
11
Authors Guild. What Authors Need to Know About the Anthropic Settlement
12
Susman Godfrey. Susman Godfrey Secures $1.5 Billion Settlement in Landmark AI Piracy Case
13
Bloomberg Law. Anthropic Judge Blasts Copyright Pact as Nowhere Close to Done
14
Publishers Weekly. Little Drama at Anthropic’s Settlement Hearing
15
Clark Hill. Right to Know June 2026
16
Reuters. US Music Publishers Suing Anthropic Make Their Case Against AI Fair Use