Licensed, Not Scraped: What a 20-Million-Paper AI Deal Tells Authors

A publishing deal announced today offers a useful glimpse of where the AI argument may be heading once we get beyond the shouting about scraping, lawsuits and whether machines can write.

Five academic publishers, BMJ Group, Sage, IOP Publishing, IGI Global Scientific Publishing and Wanfang Data have agreed to provide AI company Redpine with licensed access to 20 million peer-reviewed research papers. The material is intended for use by AI agents, with access handled as licensed, premium data rather than content simply harvested from the open web.

For novelists, that may sound like a story from another branch of publishing. I think it matters rather more than that. The important word here is not AI. It is licensed.

The announcement was reported today by The Bookseller.

In this article

Why the 20-million-paper agreement matters; why licensing is becoming AI infrastructure; what this could eventually mean for authors and book rights; and why using AI as a tool is still fundamentally different from asking it to become the author.

From scraping to a market

Much of the first wave of generative AI was built around a simple assumption: enormous quantities of digital material could be gathered first and the arguments about permission, ownership and payment could be dealt with later. Authors have spent the past few years living with the consequences of that approach.

This Redpine agreement points in a different direction. Valuable human-created material is being treated as something AI systems may need legitimate, structured access to. The publishers retain a commercial role. The AI company gets high-quality information. The transaction has an identifiable rights relationship behind it.

That does not magically solve every copyright problem, and academic journal rights are not the same as the rights attached to a novel. But it demonstrates something important: licensing is not an impossible obstacle to AI. It can be part of the infrastructure.

That direction is consistent with the UK Publishers Association's work on the emerging market for licensed AI content. Its 2026 report argues that licensing for text and data mining, AI training and retrieval-based systems is already developing across publishing. Read the Publishers Association report.

Why authors should care

If AI companies increasingly decide that trusted, licensed material is better than an endless legal fight over scraped content, books become part of a rights market rather than merely raw material for a dataset. That immediately raises practical questions for writers. Who controls the relevant right? Does an old publishing contract cover AI use? Does the author have to consent? How is revenue divided? Can a writer refuse?

We have already seen those questions surface around AI-related settlements and older publishing agreements. The difference with a licensing market is that they have to be answered before the content is used, not argued about years afterwards.

For independent authors, there may eventually be an opportunity here as well. We are accustomed to thinking about ebook, audio, translation and screen rights. AI access may become another right that needs to be understood, valued and deliberately granted or deliberately withheld.

Licensed information is also better AI

There is another side to this story that is easy to miss. Licensing is not only about compensating rights holders. It can improve the information available to AI systems. Twenty million peer-reviewed papers are valuable precisely because they are curated, structured and attributable. An AI agent able to reach reliable source material is potentially more useful than one trying to reconstruct an answer from whatever happened to be scraped into a model.

That matters to authors because research is one of the places where AI can be genuinely helpful without becoming the writer. I use AI to help investigate subjects, organise information, test questions, visualise ideas and manage parts of the creative business around my books. Those uses still require judgement, source checking and a human being deciding what belongs in the work.

If the next generation of AI tools can retrieve licensed, authoritative material and point back to it, that is a far more interesting proposition for a working writer than a machine simply producing confident-sounding prose.

AI-assisted is not AI-authored

This is also why I keep separating AI assistance from AI authorship. A system that helps me locate research, organise notes, check continuity, develop marketing or visualise a setting is supporting my work. The characters, voice, scenes, emotional choices and final prose remain mine.

Asking a machine to generate the novel and then putting a human name on the cover is a different creative act. Publishing will get into trouble if it treats those two things as identical. Sensible rules should protect human authorship without pretending useful technology has no legitimate place around it.

The same principle applies on the other side of the equation. Authors should be transparent when AI materially generates the work being sold as their own. AI companies and publishers should be equally transparent when human-created work is being used to create commercial value.

A more mature AI economy

The most encouraging thing about today's deal is that it looks less like the wild west and more like a market. Permission, access, quality and payment are becoming engineering and commercial questions rather than afterthoughts.

There will still be disputes. There will still be difficult arguments about historic training data, fair use, contracts and who is entitled to license what. But if AI is going to become permanent infrastructure in publishing, then licensed access is a much healthier foundation than assuming everything available digitally is free for the taking.

For authors, the practical lesson is straightforward: know your rights, keep your contracts and reversions organised, and watch the licensing market closely. The next important AI clause in a publishing agreement may not be about whether you used AI to write the book. It may be about whether somebody else can license the book to an AI system.

What do you think?

If AI companies are willing to pay publishers for high-quality licensed content, should authors have an explicit right to approve and share in any deal involving their books?

About Rob Frankson

Rob Frankson is the author behind 121 Minutes and the Near Galaxy Saga, writing original science fiction while exploring how emerging technology is changing the practical business of being an author. About Rob Frankson

Join / Subscribe to 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.