When AI Pays for the Book: Will the Author Actually Get the Money?

The AI copyright argument has spent much of the past few years circling one big question: were writers' books used to build commercial AI systems without permission? This week, a more old-fashioned publishing question has moved to the foreground: when money finally comes back, who actually gets paid?

The Authors Guild is pressing publishers over the distribution of proceeds connected with Anthropic's copyright settlement, particularly where books have been out of print for years and the practical ownership of rights is muddy. Publishers Weekly reported on 28 September that the Guild is asking publishers to relinquish claims on some long-out-of-print works where authors may otherwise lose part of the payout.

That sounds like a contractual dispute at the edge of a very large AI story. I think it is more important than that. It exposes something writers need to understand before the next wave of AI licensing arrives: owning the creative work, controlling particular rights and receiving the money generated by those rights are not always the same thing.

In this article

• Why the Anthropic settlement has reopened an old publishing-rights problem

• Why out-of-print does not necessarily mean rights have cleanly reverted

• What AI licensing could expose in traditional publishing contracts

• Why I separate AI-assisted author work from AI-authored fiction

• What authors should start checking now

The money has found the paperwork

The Authors Guild's own statement is blunt about the problem. It says publishers should relinquish claims on long-out-of-print books in appropriate cases and argues that authors should receive money for works publishers failed to register. The Guild also says it has heard from authors whose books have been unavailable for decades while publishers are still claiming a share of settlement proceeds. The Guild's statement frames this as an issue of fairness between authors and publishers as well as a response to AI companies' use of books.

There is an important distinction here. The settlement concerns alleged copyright infringement involving the acquisition of books from pirate sources; it should not be lazily reduced to a court declaring that all AI training on books is illegal. The legal questions around training, fair use, acquisition and licensing remain more complicated than that.

But the payment dispute tells us something practical. AI has collided with publishing contracts written for a world in which nobody imagined that a novel might become valuable as training material for a large language model.

An author's rights are becoming a live business asset

For years, rights clauses could feel remote to an independent-minded writer. Print rights, audio rights, translation, territories, reversion clauses and subsidiary rights were important, certainly, but they were not something most authors expected to collide with artificial intelligence.

That has changed. If publishers, collective licensing organisations or technology companies build legitimate markets for access to books, then the wording of old contracts suddenly matters. Who controls machine-learning or data-use rights if those words never appeared in the original agreement? Does a broad digital-rights clause cover them? What happens when a book has effectively disappeared from sale but formal rights reversion was never completed?

Those are legal questions, and individual authors should take professional advice where money or rights are at stake. But there is a simpler lesson we can act on without pretending to be lawyers: know what you signed, know which rights you retained, and keep the paperwork.

This is not an argument against useful AI

My own position on AI remains deliberately practical. I write my books. The characters, worlds, scenes and final prose are my responsibility. I do not regard asking a machine to manufacture a novel and putting my name on it as equivalent to authorship.

At the same time, I use AI around the work. Research support, visualisation, organisation, marketing, testing ideas and the administrative side of running a creative business are legitimate tools in my process. For me, AI-assisted is not the same as AI-authored.

That distinction matters here because writers do not need to become anti-technology to insist on sensible rights. I can believe AI is genuinely useful to an author and still believe that a commercial company wanting access to a writer's finished books should operate through transparent, comprehensible arrangements.

The next AI fight may be about contracts, not models

The first phase of the argument was dominated by scraping, datasets and training. The next phase may be less dramatic but more consequential for working authors: licensing terms, auditability, rights reversion and payment.

If AI companies increasingly decide that licensed material is strategically safer or simply better business, publishers will have something new to sell. Authors need to know whether the contract lets the publisher sell it, whether consent is required, how revenue is divided and whether a writer can decline.

This week's dispute is therefore a useful warning. A new technology can create a new revenue stream, but the money will still travel through old agreements. And old agreements are only as clear as the records behind them.

What I would check now

• Find your publishing contracts and rights-reversion correspondence rather than assuming the publisher's records are complete.

• Check which digital, subsidiary and future-format rights you actually granted.

• For out-of-print books, establish whether rights formally reverted or merely stopped being exploited.

• Keep AI use in your own author workflow transparent and proportionate: use the tool without surrendering authorship.

• Treat any future AI-training licence as a rights transaction, not a harmless technical permission.

The interesting question for authors is no longer only whether AI companies should pay for access to books. We also need to ask what happens after they do.

If money is generated from the use of your work, do you know who controls the right being sold — and who is entitled to the cheque?

About Rob Frankson

Rob Frankson is the author behind 121 Minutes and the Near Galaxy Saga, writing science fiction while exploring how AI can support, rather than replace the working author. About Rob Frankson and 121 Minutes.

Join 121 Minutes

AI & Editorial Transparency

AI & The Author is edited and published by Rob Frankson. Artificial intelligence is used to assist with news research, initial drafting, content organisation and supporting imagery. All articles are reviewed and, where necessary, edited by Rob Frankson before publication. The opinions, editorial position and final decision to publish remain the author's.