Analysis
Authors who thought Anthropic's $1.5 billion copyright settlement meant a straightforward check are instead getting notices that someone else has also filed a claim on their book, TechCrunch reported on September 6. The settlement, covering more than 482,000 titles Anthropic used pirated copies of to train Claude, was finalized in July at roughly $3,000 per book -- among the largest AI copyright payouts to date, and the case Pulse covered at final approval.
Who's Claiming What
The mechanics, per the Authors Guild's own explainer and TechCrunch's reporting:
“- If the book is self-published, or the publisher let the rights revert by allowing it to go out of print: the author is entitled to the full payment.”
- If a book is still in print with its original publisher: the payout splits 50/50 between author and publisher, mirroring the royalty structure in most trade contracts.
- If the book is self-published, or the publisher let the rights revert by allowing it to go out of print: the author is entitled to the full payment.
- Literary agents: some agencies are separately filing claims on titles they sold, even though agents are not rightsholders and typically take a commission on royalties rather than owning a stake in the underlying copyright.
Author Courtney Milan was among the most vocal critics of the agent claims on Bluesky this week, arguing agents have no legal basis to claim a share of a copyright settlement they were never party to as rightsholders.
Not Necessarily Bad Faith
Authors Guild CEO Mary Rasenberger told TechCrunch she doesn't believe publishers are "specifically trying to screw any author over," and instead described the dispute as the predictable result of decades of inconsistent rights reversion clauses, out-of-print determinations that were never formally documented, and a settlement claims process that had no historical template to follow. Book contracts written in the 1990s and 2000s simply never anticipated a scenario where a third party would owe a lump sum tied to the book's existence rather than its sales.
That is the structural problem underneath the fight: settlement administrators are applying modern royalty-split logic to an event -- unauthorized AI training -- that traditional publishing contracts never contemplated. There is no clause in most legacy contracts that says who gets paid if someone else infringes the book outside the normal sales channel.
The Counterweight
It is worth separating outrage from harm here. The 91% claim rate the settlement administrator has already logged suggests the process is functioning at scale even amid the disputes, and competing claims on a title do not necessarily mean an author gets nothing -- in most cases it means a negotiated or administrator-determined split rather than a denial. Anthropic itself is not a party to these individual disputes; the company already paid the $1.5 billion into a settlement fund and has no further role in how it's divided among claimants.
Why This Matters Beyond Anthropic
Anthropic's settlement was widely described as the template other AI companies facing copyright suits -- including OpenAI, which is fighting the Justice-Department-backed New York Times case Pulse has covered extensively -- would eventually have to match. If the payout process itself becomes a public mess of competing claims, it raises the administrative cost every future AI copyright settlement will need to budget for, and gives plaintiffs' firms a reason to demand cleaner claims mechanisms up front in the next negotiation.
For publishers and agents, the dispute is also a preview of a negotiating fight that hasn't happened yet: as more AI training settlements get negotiated, expect boilerplate publishing contracts to start explicitly addressing who owns the right to an AI-training payout, the same way film and TV contracts eventually built out streaming residuals language after years of ad hoc disputes.
