Authors are contesting how the Anthropic settlement gets divided
Publishers and literary agents have made claims on the fund. The writers whose books were used say the money should reach them, not the intermediaries who licensed the work.
Authors are pushing back as publishers and literary agents assert claims on Anthropic's copyright settlement, according to TechCrunch. The dispute is no longer about whether the training was lawful. It is about who owns the compensation for it.
That shift is worth marking, because it is the first time an AI copyright matter has advanced far enough for the money to become the fight. Every other case in the queue is still arguing liability.
Why the ownership question is genuinely unclear
A standard trade publishing contract grants the publisher specified rights — print, ebook, audio, translation, subsidiary rights, sometimes broad language about formats "now known or hereafter devised." It was drafted for editions, adaptations and licensing, not for a use in which the text is consumed as training data and never reproduced.
So the question is whether ingestion for model training falls inside a grant written before the use existed. Publishers point to the catch-all language. Authors argue that a clause anticipating future formats anticipates future ways of reading a book, not a use in which no reader ever encounters the text at all.
Agents' claims rest on commission entitlements attached to rights income, which depends on the same answer.
No court has construed a publishing contract against this use, which means the settlement fund is being divided on the basis of an unresolved contractual question that the settlement itself was structured to avoid.
The precedent being set quietly
Whatever allocation emerges will function as a template. It is the first substantial AI copyright settlement to reach distribution, and the next dozen will be negotiated against it.
If publishers take the majority, the working economics of AI training compensation are settled in favour of rights aggregators, and individual authors receive a residual. If authors take the majority, every publishing contract signed in the last thirty years becomes a candidate for renegotiation over a right nobody knew they were trading.
The parties negotiating this are not equally resourced. Publishers and agencies have counsel engaged on the allocation. Individual authors are represented through class mechanisms, if at all.
What it means for the labs
For Anthropic, the allocation dispute is a nuisance that follows a resolved problem, and the timing is poor: the company's public S-1 is expected in late September, and an unresolved distribution fight is exactly the kind of item that becomes a risk factor.
For the industry, the more useful lesson is about the shape of a settlement. Anthropic settled and now watches claimants argue over the fund. OpenAI and Microsoft are litigating — with a Justice Department statement of interest supporting their fair use position filed on 2 September, and new suits from the Seattle Times and Newsday filed on 5 September.
One strategy converts an open legal question into a fixed cost and a distribution problem. The other keeps the question open and the exposure unbounded. The DOJ filing has just made the second look better than it did a week ago, which is the sort of thing that makes settlements harder to reach for everyone still in the queue.
Runs the newsroom. Rename this profile in the studio to your own byline.
Related
Every weekday, the AI stories that moved money or shipped code.
No cross-posting, unsubscribe anytime. See all newsletters