Who Owns the Words Behind AI? A UK Publishing Dispute

9 hours ago

5 min read

Who Owns the Words Behind AI? A UK Publishing Dispute

AI & The Author | Friday 6 November 2026 | News reviewed: preceding 24 hours

Quick shortcuts

About Rob Frankson

Subscribe to 121minutes

A dispute over AI, educational publishing and permission

A copyright argument in Britain has moved from abstract policy into a classroom resource dispute. Reporting published on Thursday 5 November says the Publishers Association has raised concerns that Oak National Academy may have used educational publishers’ copyrighted resources to develop an AI teaching tool without permission. Oak disputes the suggestion that infringement has been established and says the association has not identified specific materials that prove its claim.

For authors, this is an important distinction. A public allegation is not a legal finding. But the disagreement raises a question that matters far beyond schools: when material is made available for one purpose, who may subsequently use it to develop artificial intelligence? The answer can depend on the original licence, the nature of the copying and the precise technology involved.

What happened in the past 24 hours?

Schools Week reported the disagreement on 5 November. The publication described concerns about resources supplied by publishers during the pandemic and their possible subsequent use in an AI tool. Oak challenged the accusation, saying the Publishers Association had not pointed to specific infringed content. Neither position should be treated as a court ruling, and the publicly available reporting does not establish that any particular book or resource was unlawfully used.

The report is relevant because educational publishing is both a creative business and an information industry. Publishers pay writers, editors, designers and subject specialists to produce reliable material. AI systems may benefit from that investment. The argument is not simply whether software is useful; it is whether permission and commercial terms travel with the content.

The difference between access and permission

One of the most persistent misunderstandings in AI debates is that content being accessible means it can be used for any purpose. An author might permit a school to display a worksheet, a library to lend a book or a reader to download an ebook. None of those permissions automatically answers whether a technology provider may copy the work into a training collection or make it available through a separate AI product.

Oak’s published licensing explanation offers an important qualification. Its 2023 statement says most new Oak curriculum resources are shared under the Open Government Licence, while third-party copyrighted material remains protected and outside that general licence. Oak’s copyright notice likewise says it seeks permission or relies on applicable legal exceptions when using material belonging to others. Those statements describe policy; they do not independently resolve the new allegation.

For independent writers, the practical lesson is to read the permitted-use language rather than relying on broad descriptions such as open, free or educational. A licence can authorise one activity and exclude another. Equally, an allegation that an AI tool used protected material must be supported by evidence about which material was used and how.

Why publishers want an AI licensing market

The Publishers Association has argued consistently that commercial licensing is the appropriate way to use publishers’ content in AI systems. Its March 2026 report described an established market for text and data mining and growing demand for licences supporting retrieval-augmented generation. That technology lets an AI assistant retrieve material while producing an answer, rather than relying only on information absorbed during model training.

Those two processes deserve separate scrutiny. Training may involve analysing large collections to develop a model. Retrieval may involve finding and presenting particular passages at the moment someone asks a question. Rights, permissions and potential economic effects can differ. Treating every use of AI as one undifferentiated activity makes it harder to negotiate fair terms or identify an actual breach.

The association’s broader position is that the UK should require transparency from AI developers and protect a functioning licensing market. AI companies and public bodies, by contrast, often emphasise the public value of efficient access to information and the possibility of lawful uses under licences or statutory exceptions. The disagreement is about the boundary, not whether teachers or authors should benefit from technology.

A second warning: disclosure is becoming part of authorship

Another story published on 5 November illustrates a different aspect of trust. The Guardian reported that French literary prize winner Philippe Jaenada acknowledged limited uses of AI while writing his award-winning novel. His account distinguishes between using AI for particular research or deliberately incorporated passages and handing over the substance of a novel to a machine. The episode also revived debate about whether writers are judged consistently when AI assistance becomes public.

That matters because disclosure can be more informative than a simple yes-or-no label. A spelling check, an alternative outline, a generated paragraph and a wholly machine-written book are not the same creative act. Readers deserve accuracy about substantial contributions, but a blanket suspicion of every tool risks obscuring what an author actually created.

Three actions for working authors

First, keep a clear rights record. Save contracts, licences, publication dates and manuscript versions. If you supply material to a school, publisher, platform or technology company, note what use you authorised. Documentation is more useful than an argument conducted from memory months later.

Second, make your own AI policy specific. Explain whether tools help with research discovery, administration, editing suggestions, marketing or imagery. Where AI contributes substantive prose or artwork, check contractual requirements, rights and disclosure expectations before publication. The goal is not a fashionable badge; it is a statement readers can understand.

Third, verify claims at their source. In the Oak dispute, read both the contemporary reporting and Oak’s own licensing statements. Separate a publisher’s concern from a demonstrated infringement. The same discipline applies when an AI assistant supplies a quotation, a legal conclusion or a statistic: follow the reference, check the date and confirm the original document.

My view: permission and provenance should be ordinary practice

I write science fiction because technology creates possibilities that are worth exploring. But the future of publishing cannot be built on vague assurances about where words came from. A useful AI system should be able to explain what content it may access, on what terms, and what it has done with that content. Authors should be equally clear about the human decisions behind their books.

This week’s British dispute does not establish wrongdoing. It does show why provenance is becoming a practical business issue. The best outcome would be evidence-led scrutiny, workable licences and systems that support education without treating creative labour as an unlimited free resource. Authors do not have to reject AI to insist on those standards.

Sources and further reading

Schools Week — reporting on the Oak copyright dispute, 5 November 2026

Oak National Academy — licensing and third-party rights policy

Oak National Academy — copyright notice

Publishers Association — AI licensing market report, March 2026

Publishers Association — artificial intelligence policy

The Guardian — French literary prize winner’s AI disclosure, 5 November 2026

About Rob Frankson

Rob Frankson writes original science fiction in the Near Galaxy Saga and explores AI, creativity and the business of writing at 121minutes.com.

Editorial transparency

AI & The Author is edited and published by Rob Frankson. Artificial intelligence may assist with news research, initial drafting, organisation and supporting imagery. All articles require human review before publication; editorial judgement and the decision to publish remain with the author.