Publishers Want AI Licensing Deals. But Who Speaks for the Author?

Something important is changing in the argument about artificial intelligence and creative work. For a long time the conversation centred on whether technology companies could take material from the internet and use it to train their systems. Now the argument is moving into the commercial world: who can license that material, on what terms, and who actually receives the money? That is a much more practical question for a working author. It is also one we cannot afford to leave entirely to publishers, technology companies and lawyers.

This week, renewed discussion of publisher licensing and new AI content partnerships makes the question especially timely. Reuters reported on 15 October that CNN, Axios, the Financial Times and other news organisations have joined Google's expanded AI news programme. Publishers are licensing content for AI-powered experiences in return for compensation, while attribution and traffic back to the original source remain important points of negotiation. The agreements concern news content rather than novels, but they offer a useful glimpse of how the wider market may develop.

The deal is not the whole story

A licensing agreement sounds reassuring. Rather than taking creative material without permission, a company pays for access. That is progress in principle, but a headline announcing a deal tells us very little about the rights of the individual creator. What material is included? Who owns or controls the relevant rights? Can the licence be revoked? Is the payment based on training, retrieval, summarisation or some other use? And if a writer's work helps make the deal valuable, how much of that value reaches the writer?

The distinction between news publishing and book publishing matters here. A newspaper may own or control certain rights to the articles it commissions. An author signing a book contract usually grants a defined package of rights, while retaining others. We should not assume that because a publisher has a book in its catalogue it can license every possible AI use of that book. The contract, applicable law and specific rights granted remain decisive.

Britain is already building a licensing market

The UK's Publishers Association made the commercial direction clear in its March 2026 report, Content Superpower: UK Publishing and the AI Licensing Market. It described a growing market for licensed publishing content, including arrangements for training and retrieval-augmented generation, where AI systems consult material to answer questions. Publishers' Licensing Services has also developed an industry-led collective licensing initiative, inviting publishers to opt in their content. Authors' Licensing and Collecting Society is involved in the author-facing side of that conversation.

Collective licensing can be useful. Individual authors rarely have the time, leverage or resources to negotiate with every technology business. An organised system may reduce transaction costs and make legitimate use easier to track. But convenience is not a substitute for permission. If the system is to work for authors, it needs transparent terms, a clear record of which works are covered and a credible mechanism for distributing revenue.

It also needs a way to say no. An opt-in system has a different ethical foundation from a system that assumes consent until an author manages to find an obscure opt-out form. Writers should know what they are agreeing to and what uses their consent covers. Licensing for a searchable research tool is not automatically the same as licensing a novel to train a model that generates prose in a similar market.

Training, retrieval and generated output are not interchangeable

This is where the language around AI becomes unhelpfully broad. Training is the process of using material to develop or refine a model. Retrieval may involve consulting a licensed source when responding to a question. Generated output is what the user eventually sees. These activities can overlap, but they are not the same. A contract that merely says 'AI rights' may be too vague to protect anyone's interests.

For an author, the commercial consequences differ. A reader who asks a service for a citation from a licensed research publication may be directed to the original source. A reader who asks for a synthetic novel in a particular style may never encounter the original author's work at all. The questions of attribution, substitution, payment and creative identity change with the use. Good licensing should recognise those differences rather than bundle everything together.

Assistance is not authorship

I also think we need to preserve another distinction that is too often lost. Using AI to organise research, test a marketing approach, visualise a fictional location or manage the administrative side of an author's business is not the same as having a machine write the fiction and presenting it as one's own. There can still be questions about accuracy, confidentiality, image rights and the origin of training material. But the central creative choices of a story belong to the writer who makes them.

The more we muddle these categories, the harder it becomes to write sensible rules. A blanket declaration that all AI use is identical does not help a novelist decide whether to use an organisational tool. Equally, calling every generated passage 'assistance' obscures the difference between supporting a writer and replacing the writing itself. A credible industry policy should identify what was done, who did it and what rights were involved.

What should authors ask before agreeing?

I would start with five questions. Which exact works and rights are being licensed? Is the use limited to training, retrieval, summaries or other defined purposes? What compensation is payable, how is it calculated and who receives it? What evidence will authors receive about usage and distribution? Finally, can consent be withheld or withdrawn, and what happens to existing copies or models if it is? These are not hostile questions. They are the basics of a commercial arrangement.

The Authors Guild's model AI contract clauses are worth reading because they make explicit that authors should retain control of AI-related rights unless they choose to grant them on negotiated terms. The Guild also addresses publisher uploads of manuscripts to commercial systems and the importance of permission and safeguards. Although US model clauses are not a substitute for UK legal advice, they demonstrate the level of specificity that is possible.

The market needs the writer at the table

The most encouraging aspect of the current licensing debate is that the value of human-created content is finally being acknowledged in commercial terms. The danger is that the people who created that value could be the last to learn how it was sold. I do not want to see authors reduced to a line item in a negotiation conducted elsewhere. I want a market in which writers understand the choices, retain meaningful control and share fairly in the returns.

AI licensing is not going away. The question is whether we build the rules around the convenience of the largest companies or around the rights of the people whose work makes those systems useful. For me, that is the story behind the headlines about new partnerships. And it is the part authors should be watching most closely.

Sources and further reading

Reuters, CNN signs multi-year AI partnership with Google (15 October 2026): https://www.reuters.com/business/media-telecom/cnn-signs-multi-year-ai-partnership-with-google-2026-10-15/

Publishers Association, Content Superpower: UK publishing and the AI licensing market (3 March 2026): https://www.publishers.org.uk/publications/content-superpower-uk-publishing-ai-licensing-market/

Publishers' Licensing Services, Generative AI Licensing: https://pls.org.uk/generative-ai-licensing/

Authors Guild, AI-Related Model Publishing Contract Clauses: https://authorsguild.org/advocacy/artificial-intelligence/ai-model-clauses/

About Rob Frankson: https://www.121minutes.com/about

Join / Subscribe to AI & The Author: https://www.121minutes.com/subscribe

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.