Authors Challenge Anthropic Settlement Claims as Publishers and Agents Seek Payment Cuts

Authors expecting payments from Anthropic’s $1.5 billion copyright settlement are now reporting a new problem: publishers and, in some cases, literary agents are also claiming a share of the money. The dispute matters because the settlement is meant to compensate writers whose books were allegedly used without permission to train AI systems, and any mistaken allocation could delay or reduce those payments.

The issue emerged this week as authors began receiving notices that third parties had filed claims on books covered by the class-action deal. Some writers say the claims appear to be based on outdated rights records, while others argue that agents have no ownership stake at all and should not be involved in the payout process.

What is happening with the Anthropic settlement?

Authors are flagging what they say are incorrect or unexpected claims on the money they are due under Anthropic’s copyright settlement. The underlying case was resolved after a court found that training AI systems on copyrighted works can fall under fair use, but that using pirated copies is a separate legal problem. Final approval of the settlement came in July, clearing the way for payments to begin.

Under the agreement, authors of nearly 500,000 titles are set to receive $3,000 for each pirated work covered by the class action. In many cases, though, the payout may not go directly to the writer alone. If a book remains in print with a traditional publisher, the settlement is supposed to split the payment evenly between author and publisher. If the book is self-published, or if the rights have reverted after the title went out of print, the full amount should go to the author.

Why are authors pushing back?

Authors say the claims they are seeing do not match the settlement’s rules. Some say publishers have asserted rights over books that should have already reverted back to the writer. Others say publishers are seeking more than half of the payment even where the agreement appears to limit them to 50%.

For many writers, the surprise has been as unsettling as the financial stakes. These are not minor administrative quirks when multiplied across hundreds of thousands of books and potentially thousands of individual claims. A mistaken allocation can mean weeks or months of delay, appeals, or direct disputes between an author and a publisher.

Mystery and thriller author April Henry said she was stunned to learn that HarperCollins had allegedly claimed a book she said reverted to her many years ago, and she also received an alert indicating the publisher had been listed as her employer, something she said never happened.

Henry’s complaint reflects a broader concern among writers: that the system handling the settlement may be relying on incomplete or outdated metadata rather than current rights records. In the world of publishing, those records are often fragmented, especially for older titles, books that have gone out of print, or contracts that have changed hands over time.

How are publishers defending the claims?

Some publishing companies appear to be treating the problem as a data issue rather than a deliberate attempt to take money from authors. Writers Beware, the long-running author advocacy blog run by Victoria Strauss, has said complaints seem to fall into two main categories: publishers claiming money on works they no longer control, and publishers requesting the full payout when they should only receive half.

Strauss has said she is hesitant to assume bad faith when poor recordkeeping could explain at least part of the confusion. She also noted that some publishers have already said the claims were filed in error and that they have asked Anthropic to correct them.

Authors Guild CEO Mary Rasenberger made a similar argument in comments to The New York Times, saying she does not view the situation as a coordinated attempt by publishers to seize author compensation. Instead, she described it as the predictable outcome of weak records and a settlement framework that can be hard to navigate.

What makes the settlement process confusing?

The settlement distinguishes between books that are still commercially active and those whose rights have reverted to the author. That sounds straightforward, but in practice it can be difficult to verify. A book may have been sold, licensed, reissued, or allowed to lapse years earlier, and not every participant in the chain may have updated records.

There is also a key date that matters for authors seeking to claim a full payment. If the rights to a book reverted before August 10, 2022, the author may be entitled to the entire settlement amount for that title. If the reversion happened after that date, the split may be different. That detail alone can make disputes difficult to resolve quickly.

Are literary agents also claiming a cut?

Some writers say yes, and that has drawn especially sharp criticism. Strauss reported complaints that several literary agencies have also made claims against settlement payments. That surprised her because agents typically broker deals and manage relationships, but they are not usually the owners of the underlying book rights.

In other words, an agent may have earned commissions on the original book deal, but that does not necessarily create a legal entitlement to money from a copyright settlement tied to the work itself. That distinction is now at the center of frustration among authors who fear that too many middlemen are appearing in a process intended to compensate creators.

Author and legal scholar Courtney Milan, writing on Bluesky, said she believed some agents were trying to claim percentages of the Anthropic settlement and argued bluntly that they should stop.

Milan’s reaction underscores the emotional dimension of the dispute. For many authors, the settlement is not just another royalty stream; it is compensation linked to allegations that their books were used in ways they never authorized. Any appearance that non-rightsholders may be reaching for a share can quickly become a flashpoint.

How big is the problem?

No one yet knows whether the disputes involve a handful of mistakes or a broader failure in the claim system. Strauss has cautioned that the complaints she has received represent only a small sample of the overall process. Still, she also said the volume and repetition of the reports are notable.

According to Strauss, the same kinds of errors are appearing over and over, which raises the possibility that the problem is more systemic than a few isolated glitches. That distinction matters because isolated errors can be fixed one by one, while systemic mistakes may require a broader review of how claims are being matched to titles and rights holders.

The central challenge is scale. The settlement covers nearly half a million books, and each title may have a different rights history, publisher arrangement, or chain of ownership. Even a small error rate can affect many people when the dataset is this large.

Why this settlement matters beyond one company

This dispute goes well beyond Anthropic. It is becoming a test case for how the publishing industry, authors, and AI companies will handle ownership, licensing, and compensation when large language models are built using books and other copyrighted material.

The settlement also highlights a deeper problem in the AI era: legal resolution does not always mean operational clarity. A court can decide what kind of training is lawful in general, but turning that ruling into actual payments requires clean rights data, accurate contract histories, and a process that every party trusts.

If those systems break down, the result is likely to be conflict not only between authors and AI firms but also among authors, publishers, agents, and claims administrators. That could slow future settlements and shape how rights holders approach any new licensing deals with AI developers.

What authors should do next

Authors who believe their claims are wrong are being urged to review the allocation details carefully and dispute any incorrect entries through the process described by the Authors Guild and other advocacy groups. The exact steps may vary by title, but the first priority is documenting rights reversions, publishing status, and any contract terms that affect ownership.

In practice, the best evidence may come from old publishing agreements, reversion notices, out-of-print records, and correspondence showing when a book returned to the author. Where that documentation exists, it may help writers challenge claims that publishers or agents cannot substantiate.

• Check whether the book was still in print at the time of the claim.
• Confirm whether the rights had reverted to the author before the settlement cutoff date.
• Gather contracts, reversion letters, and publishing correspondence.
• Compare the claimed payout split against the settlement rules.
• File a dispute promptly if the claim appears inconsistent.

How did the dispute reach this point?

The dispute grew out of a landmark copyright case that forced the AI industry to confront how training data is sourced. Anthropic’s settlement was meant to resolve one part of that conflict while allowing payments to move forward after court approval in July. Instead, the payout phase has exposed a second layer of tension: who exactly should be paid and how much.

That tension is not unique to Anthropic. Similar questions are likely to arise wherever AI companies settle claims tied to copyrighted works, especially when old publishing contracts, shifting rights ownership, and imperfect databases intersect. The more titles involved, the more likely it becomes that someone will be left off a claim, over-credited, or wrongly inserted into the process.

For now, the immediate story is one of authors trying to protect compensation they believe is already theirs. But the larger story is about whether the industries surrounding books can build settlement systems that are accurate enough for the scale of modern AI disputes.

Timeline of the Anthropic copyright settlement

The sequence of events helps explain why authors are now raising alarms.

What comes next?

The next phase will likely depend on how many of the disputed claims are corrected and how quickly administrators can separate legitimate rights holders from mistaken ones. If the complaints remain limited, the settlement may proceed with relatively modest friction. If the same errors keep surfacing, the case could become a warning sign for future AI copyright settlements.

For authors, the immediate goal is simple: get the correct people paid. For publishers and agents, the challenge is to prove any claim they make is backed by current rights information. And for Anthropic, the settlement will be judged not only by its size but by whether it can be administered fairly and transparently.