AI Licensing Is Becoming a Market. What Should Authors Demand?

AI & THE AUTHOR — 21 September 2026

Quick Shortcuts: The shift from argument to market • Why licensing matters to writers • AI-assisted is not AI-authored • What authors should watch next • Discussion question

For most of the last four years, the argument over books and generative AI has sounded almost binary: technology companies say large-scale training can be lawful; authors and publishers say creative work should not simply become free industrial raw material. But something more practical is beginning to emerge underneath the lawsuits and rhetoric. AI licensing is becoming a market.

That is the most important development for working writers right now because a licensing market changes the question. Instead of endlessly asking whether AI companies can use books, we can start asking under what terms they should use them, who gets paid, who gets to say no, and what information authors receive about where their work goes.

The debate is becoming a market

Publishers’ Licensing Services chief executive Tom West argued this month that the industry now needs to move from principles to workable licensing solutions. That matters. Licensing is not a philosophical compromise between writers and technology. It is one of the mechanisms copyright industries have always used to allow valuable work to be reused without pretending that the original creator has disappeared from the equation.

At the same time, the legal pressure has become very real. In July, a US federal judge approved Anthropic’s $1.5 billion settlement with authors over allegations involving pirated copies of books. The underlying litigation drew an important distinction: training itself and the way copies are obtained are not necessarily the same legal question. Then, during September, authors began disputing how settlement payments were being claimed and divided between writers, publishers and agents. Suddenly, AI rights are not theoretical. They involve contracts, rights reversions, record keeping and actual money.

Why writers should care now

For an author, a functioning licensing system could offer something the current environment largely lacks: visibility. A sensible market should be able to identify the work, identify the rightsholder, define the permitted use, record the licence and provide a route to remuneration. It should also make refusal possible where appropriate. None of that requires us to pretend AI is going away.

Publishing is already using AI in less controversial areas. Publishers Weekly reported this month that publishers are deploying it for metadata, royalty statements, proofreading, forecasting and customer service. Those uses are quite different from replacing an author or absorbing a novel into a training corpus without a clear rights framework. Treating every use of AI as identical makes it harder, not easier, to defend the parts of authorship that actually matter.

AI-assisted is not AI-authored

My own line remains straightforward. I write my books. The characters, worlds, prose and final creative decisions are mine. I am comfortable using AI around that work for research support, visualisation, organisation, marketing and the practical business of being an independent author. That is assistance. Asking a machine to manufacture the novel and then presenting the result as my writing would be something fundamentally different.

The useful distinction is not ‘AI or no AI’. It is who made the creative decisions, who is responsible for the finished work, and whether other people’s creative property was used on fair and transparent terms.

That distinction should work in both directions. Writers are increasingly expected to explain meaningful AI involvement in what we publish. A mature AI licensing market should expect comparable transparency from technology companies: what kinds of works are licensed, what uses are permitted, how rights are tracked, and how creators participate economically.

Licensing will not solve everything

There are difficult questions ahead. Publishers may control some rights while authors retain others. Older contracts were never written with model training in mind. Rights may have reverted. Different territories have different copyright rules. Collective licensing may be practical for millions of works, but authors will reasonably want to know what has been licensed on their behalf and whether they can opt out.

There is also a danger that licensing becomes merely a legitimacy sticker: a small payment in exchange for enormous downstream value. A healthy market therefore needs more than permission. It needs terms that creators can understand, reliable records, meaningful remuneration and enough competition that authors and publishers are not negotiating with a single unavoidable buyer.

A more useful argument for authors

For writers, I think this is a better place to put our energy than trying to turn the clock back. Generative AI exists. Publishers are using AI tools. Readers are encountering AI-generated material. Courts are gradually defining legal boundaries. The question is what sort of creative economy we build around those facts.

Licensing offers the possibility of an answer that does not require choosing between technological development and authors’ rights. AI companies can gain access to high-quality, legally sourced material. Publishers can develop new rights businesses. Authors can retain a place in the value chain instead of discovering after the event that their work helped create somebody else’s product.

That is not automatically a good deal. But it is at least a deal and something that can be examined, negotiated and improved.

What I would watch next

I would watch three things closely: whether licensing schemes tell authors exactly what rights are being granted; whether payments reach the correct rightsholders when publishing rights have changed or reverted; and whether licensing remains voluntary enough to preserve genuine author choice. Those details will tell us whether the emerging market is really a creator economy or simply a more organised extraction system.

The question

If an AI company offered to license one of your books for model training under transparent terms, with a clear payment and the right to refuse, would you take the deal — and what would you need to know before saying yes?

Sources & further reading

Publishers’ Licensing Services — AI licensing landscape and emerging market. Reuters — US judge approves Anthropic’s $1.5 billion copyright settlement. Publishers Weekly — Publishing’s AI Reckoning. Jane Friedman — AI and Publishing: FAQ for Writers.

About Rob Frankson

Rob Frankson is a science-fiction author and creator of the Near Galaxy Saga. Through 121 Minutes and AI & The Author, he explores writing, publishing, technology and the practical realities of building a creative career while keeping human authorship at the centre of the work.

Join 121 Minutes

Follow AI & The Author and the wider 121 Minutes project for continuing discussion about books, publishing, science fiction and the changing creative landscape. Visit 121minutes.com to read more and sign up.

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.