In June I did the per-book math on Anthropic's $1.5 billion settlement with book authors: "Do the division on the full $1.5 billion across the works actually in the class and you land a hair above three thousand — call it roughly $3,100 a book once the arithmetic settles." I closed that piece with a forecast: "The number, for now, is about three thousand dollars a book. It will change."
It has — though not in the way that piece imagined. On September 2, 2026, class counsel's status report put the first distribution at approximately $2,203.56 per claimed work, after fees, service awards and expenses, and before Anthropic's final $450 million payment. And each book's share still has to be divided among everyone who claimed it. Often that means the author and the publisher. Some publishers claimed all of it (several have since called that a mistake), and in a number of reports a literary agency turned up as a claimant, in many cases by error.
So let me own what the June arithmetic left out: it divided the money by books, never a book between the people who claim it. That second division is where the fight has moved — to authors and their own publishers.
If your book is on the list, what matters now is the percentage beside your name on your notice — and the administrator has asked for anything showing it is wrong within 60 days of the day you received it.
From about $3,000 to about $2,200, in two steps
Why the drop? Two reasons, both written into the deal from the start.
First, $3,000 was always the gross. When the court approved the settlement on July 20, 2026, it quoted its own earlier order: class members "get a per-work award of about $3,000, less costs and fees." The court awarded class counsel $101,561,111, and 70% of that comes out of this round, along with roughly $21 million for expenses, three service awards and a reserve for future costs.
Second, this round spends only the money that has arrived. As of the September 2 report, 70% of the $1.5 billion, plus interest, was in the fund; Anthropic's last $450 million, plus interest, is due no later than September 25, 2027. What remains after the deductions, roughly $991 million, is spread across a final count of 449,731 claimed works: "this initial distribution should amount to $2,203.56, subject to adjustments," the report says (interest is still accruing, for one thing).
So $2,203.56 is a first round, from a fund the settlement agreement makes non-reversionary (none of it goes back to Anthropic). And it belongs to the work, so it still has to be split: under the default described below, each side's half of this round is roughly $1,100, by my arithmetic.
Why does a publisher get any of it?
Because, in this settlement's terms, a traditionally published book often has two owners: the publisher as "legal owner," holding the exclusive right to publish by contract, and the author as "beneficial owner," who handed that right over for royalties. Penguin Random House's author FAQ calls them separate class members, each filing its own claim.
When both claim a trade or university-press book, the settlement's FAQ says, the payment "by default, will be split 50-50 between authors and publishers." The reason is ordinary contract language: "Many publishing contracts provide that copyright infringement recoveries are shared equally between authors and publishers."
The court-approved plan calls that default "non-mandatory." Any claimant can depart from it by checking a box and submitting documents, publishing agreements included. Two groups sit outside it. A sole owner (an author whose rights reverted or whose contract ended, a self-published author, or a publisher that commissioned a work for hire or was assigned all the rights with no royalties owed) is "entitled to 100% of the settlement proceeds allocated to the work." And education works (textbooks and other titles from education publishers) get no default at all — their contracts are, in the FAQ's words, "sufficiently diverse" — so the claimants work out the split from the contract instead.
So the default is a starting point anyone can move, with paper. Remember that last phrase.
What authors found when they logged in
So far, so good: a default, an exit for anyone with documents, and a clear rule for sole owners. Then the notices went out.
Between September 2 and 4, the court-appointed administrator, JND Legal Administration, emailed claimants a notice showing who else had claimed each of their works, and for what share.
Victoria Strauss, co-founder of Writer Beware, the writers' watchdog, said most of the reports reaching her fell into two piles: "publishers making 50% or 100% claims on rights-reverted works (where they should receive no share), and publishers making 100% claims on in-print works (where their share should be 50%)." Some of those in-print claims, she noted, "may reflect the publisher's ownership of copyright." The thriller writer April Henry posted that HarperCollins had claimed one of her books that "reverted back at least 17 years ago," as TechCrunch reported. She told The New York Times she didn't think Harper "was deliberately trying to cheat."
By September 8, Writer Beware was reporting that HarperCollins/Harlequin "has corrected the claims on rights-reverted books" and that Kensington was doing the same. Kensington, like some other publishers, had misunderstood how to enter the percentages, Publishers Weekly reported; its chief executive, Steve Zacharius, said, "We had always planned on the author getting their 50% share." The Authors Guild relayed that some publishers had told the administrator they "mistakenly selected a 100 percent allocation rather than the default option," and the administrator began updating them.
As for agents, five authors told Strauss early on that six agencies had claimed 15% to 25%; her running list of reported agencies had grown to 14 by October 3. "Agents are not rightsholders in the books they sell," she wrote, and by September 8 she was reporting that "many of the apparent claims by agents are errors."
Education works are a different case, and a harder one. Authors with Wiley and Pearson told Strauss their publishers were claiming between 75% and 90%, and she added: "This isn't necessarily a wrongful claim." Where an education contract says nothing about infringement awards, the Guild explains, many education publishers apply the royalty rate: "if the royalty rate is 10 percent to the author, educational publishers are claiming 90 percent of the award." (That would leave the author about $220 of this first round, by my arithmetic.) Wiley told the Times that allocations for its educational titles follow individual contracts; as of October 3, I could find no public statement from Pearson. The Textbook & Academic Authors Association, which advocates for these writers, says some textbook authors report publisher claims of "85-100%" and tells them: "You do not have to accept that percentage allocation."
A grab, or a filing problem?
It would be easy to write this as a heist. The authors' own advocates decline to.
Strauss writes that she is "reluctant to attribute to malice what can be plausibly explained by poor recordkeeping," and suspects that poor records, poor checking, or overworked or inexperienced staff "are responsible for many if not most of the wrongful claims on rights-reverted works." (She adds, fairly: "Not that that's an excuse.") Mary Rasenberger, chief executive of the Authors Guild, told the Times, as TechCrunch reported, that she doesn't see this as "a grab by the publishers." She told Publishers Weekly that the most common problem she sees is authors claiming 100% because they believed their rights had reverted; as PW summarized, reversion records are "often incomplete (or non-existent)."
Some publishers took the default from the start: Princeton University Press says it "submitted all Princeton University Press claims at this 50/50 split," because its contracts divide infringement recoveries equally. And the Guild itself explains the default as a reflection of industry contract norms, meant "to make the claims process speedy and efficient."
I take all of that at face value. But if bad records explain much of this, the outcome turns on who holds the records and who has to produce them. Under this plan the default stands unless someone documents a different split — so an author whose publisher still lists a long-reverted book as its own has to produce the paper to get back to 100%. "A lot of authors didn't have their reversion letters," the New Hampshire author Margaret Porter told the Concord Monitor. On September 25 the Guild asked publishers to withdraw their claims on books that have been long out of print, proof of reversion or not.
Europe has already fought over the publisher's half
So who should set the default? Europe has fought over almost this exact question for a decade, about a different pot of money, and its answer kept moving.
The money there is the "fair compensation" paid to rightholders for copying the law allows, such as photocopying. Belgium's law said it was to be "allocated in equal parts to authors and publishers." In November 2015, in Hewlett-Packard Belgium v Reprobel, the EU Court of Justice said publishers "are not among the reproduction rightholders" under the EU's 2001 copyright directive, and ruled that national law could not hand them part of the authors' compensation where they were "under no obligation to ensure that the authors benefit, even indirectly." (The no was conditional.)
Germany's Federal Court of Justice followed the EU ruling in April 2016, holding that Germany's collecting society for writers and their publishers could not keep paying publishers a flat amount of, as a rule, half its income; its money had to go "solely to the entitled parties."
Then the EU reopened the door. Article 16 of the 2019 copyright directive says "Member States may provide" that an author's transfer or license of a right to a publisher is a sufficient legal basis for the publisher to share in that compensation. Its recital 60 makes the publishers' case (they "make an investment" and "can in some instances be deprived of revenues"). Germany's law now gives publishers a share again, with a floor: the author is entitled to at least two-thirds of the income, unless the collecting society sets a different split.
Count who held the pen: in Belgium, a statute; in Germany before 2016, a collecting society's distribution plan; since 2019, each EU country; in Germany now, a statute with a floor its collecting society can adjust. In the Anthropic settlement, a court-approved plan set a non-mandatory half, contracts override it, and a court-appointed Special Master settles what the parties can't.
Europe's author-first academics, the European Copyright Society, argued in 2015 that publishers' protection "should derive either from the contracts concluded with the individual creators or by way of a related right granted by law." The American plan sits close to the first half of that sentence, which I think is the right instinct. To be clear, I am not holding Europe up as the model; its default moved three times in under four years. But a default is never neutral — it decides who has to go to court, or to the filing cabinet.
The next AI check will meet the same old contracts
What happens when the money stops being a one-off? A settlement looks backward; licensing money can arrive every year.
HarperCollins' opt-in AI licensing deal, Publishers Weekly reported in November 2024, offered $5,000 a book, "split evenly between the author and the publisher at $2,500 each," according to PW's source. The libertarian magazine Reason expects AI companies to "turn to a growing market for repositories of licensed copyrighted works."
So run it forward (this part is my extrapolation). Say it is 2031, and your 2012 novel sits in several licensed AI libraries, each paying a fee every year. Your contract never mentions AI. The money then flows by whatever default the licensing form carries: half, as Harper's deal chose; the royalty rate, as many education publishers read their contracts now; or everything to whichever party's records still show the rights. Your reversion letter is in a drawer — or it isn't. And the question this fall's portal asks you once (who owns this book, and who can prove it?) comes back every year, on every license.
Who defends the half, and who doesn't
On the left, the antimonopoly Open Markets Institute warned in November 2025 that even for a book in line for the full payout, "the author will likely have to split the settlement 50/50 with their publisher, and that's after legal fees and other costs are deducted." That is roughly what happened.
Indiana University law professor Lea Bishop went further in a formal objection: "Publishers did not file this case because these damages are not theirs to recover." The court overruled every objection. Even the Authors Alliance, an author-side group, acknowledged that "publishers are likely to benefit substantially from this settlement."
In Europe, IFRRO, the federation of reproduction-rights organizations, answered Reprobel by asking the EU to keep "fair and adequate compensation for both authors and publishers."
What does this mean for you?
If a book of yours is on the Works List, this is what the settlement's own pages and the authors' groups advise, as of October 3, 2026:
Find your notice. It came from info@anthropiccopyrightsettlement.com between September 2 and 4, or by mail later if the administrator had no valid email for you. Check your spam folder; if there is nothing, call the administrator at 1-877-206-2314. The Guild warns of look-alike scam domains: real links go to www.anthropiccopyrightsettlement.com.
Check every title, even the quiet ones. Log in through the portal link in your notice with its claim number and PIN. Writer Beware: "Don't assume, just because the page says No Disagreements, that everything is all good!" Enter your Social Security or taxpayer ID number (outside the U.S., a foreign tax ID on a W-8BEN form, per TAA) and click submit; until you do, TAA warns, "you have not completed the process."
Know your default. Trade or other non-education book, publisher holds the rights: 50/50 unless someone documents otherwise. Textbook or other education work: no default; your contract governs. Rights reverted, contract terminated, or self-published: "you can claim the entire award." Reverted after August 10, 2022 (the settlement's "Download Date")? The Guild says the publisher "may have a valid claim." Co-authors split the author's side equally.
Gather the paper first. The contract (Writer Beware says to check whether its copyright clause splits litigation recoveries equally), any reversion letter or written request for one, and royalty statements. No letter, and no royalty statement in a long time? The Guild says you may still claim 100% with a signed statement that you believe the rights reverted, though success "will depend on the case." Upload it, then email it to the publisher with a copy to the administrator.
Count 60 days from your own notice. The settlement site extended the window from 30 days to 60 and asks for supporting information "within 60 days of receiving access": around November 1, 2 or 3 for notices emailed September 2, 3 or 4, later for mailed ones. Check the date on yours. The window "does not create a deadline by which all disagreements between co-claimants must be resolved"; a work still in dispute misses the first round, and its money waits in reserve for a later one.
Write to the publisher. Each of the Big Five has a settlement mailbox, listed on the Guild's page. No contact, or no answer? Go to the administrator and class counsel. Out of print for years? Ask the publisher to withdraw its claim.
Know the last stop. If talks fail, claimants may take the dispute to Special Master Theodore K. Cheng, "at no cost to the claimants"; his decisions are "final and binding without a right of appeal." The Guild will represent members before him.
Be wary of cash-now offers. One buyer has offered authors "up to" $1,500 a book for their payments. Writer Beware says the offer "may be predatory, but it's not illegal" (assuming the buyer isn't "just a scam front"), and that it "doesn't really improve on what many authors will collect anyway."
Expect more than one round. First payments are expected on or before November 15, 2026 for works with no disagreement and complete payment information; the approximately $2,203.56 figure does not include Anthropic's final $450 million.
The lesson, as I see it
A default is a decision made in advance for everyone who doesn't show up with paper. This one is defensible: it mirrors common contract language, and the court approved it. It still decides who has to go and find the letter.
In June I also told you to keep records of authorship, since in a per-work world, you can't get paid for what you can't prove you made. That was half the advice. The fights in this settlement are over ownership, and ownership lives in contracts, reversion letters and royalty statements that too often went missing, on both sides.
My vote? Treat a reversion letter like a deed: keep it, and send copies to the publisher and the administrator. Write AI licensing into contracts in plain numbers now, rather than leaving the split to an infringement clause drafted for a different purpose. And the next time someone announces a per-book figure for AI money, ask who wrote the default before you ask how big the number is.
Your sixty days started the day your notice arrived, not the day you read this.