CruxOne
← All posts

News··9 min read

“Don’t Steal This Book”: where UK and EU copyright rules on AI training stand in 2026

The Don’t Steal This Book protest, the UK’s March 2026 AI copyright report and the EU AI Act, explained, plus what UK and EU authors can do right now.

A stack of plain hardback books with blank covers on a table at a busy book fair
On this page

As of November 2026, neither the UK nor the EU has changed the law to let AI companies train on books without permission. The UK government dropped its plan for an opt-out copyright exception in March 2026 and has not legislated since, while the EU AI Act already requires AI model makers to respect authors’ opt-outs and publish summaries of their training data. Here is what the “Don’t Steal This Book” campaign was about, what is law and what is still only a proposal, and what indie authors can do now.

What the “Don’t Steal This Book” campaign is

“Don’t Steal This Book” is a protest book with no story in it, only a list of author names, published to oppose AI companies using writers’ work without permission. It was organized by Ed Newton-Rex, a composer and copyright campaigner, and handed out at the London Book Fair, which ran from March 10 to 12, 2026.

According to downthetubes.net, the book carries the names of about 10,000 authors, and organizers gave away 1,000 free copies at the fair. Euronews named Kazuo Ishiguro, Richard Osman, Alan Moore, Marian Keyes, Malorie Blackman, Philippa Gregory and Mick Herron among the signatories. The back cover reads: “The UK government must not legalise book theft to benefit AI companies.”

The timing was deliberate. The UK government was due to publish its report on copyright and AI one week later. The same week, the Society of Authors launched a “Human Authored” logo scheme at the fair, built with the US Authors Guild, so authors can mark books written without generative AI.

What the UK government decided in March 2026

The UK government decided not to change copyright law for now. Its report on copyright and artificial intelligence, published on March 18, 2026 under section 136 of the Data (Use and Access) Act 2025, dropped the option it had favored in its consultation: a broad text and data mining exception that would have let AI firms train on works unless rights holders opted out.

In the report’s words, “a broad copyright exception with opt-out is no longer the government’s preferred way forward.” Most consultation responses opposed it. The government said it will not reform copyright law until it is confident changes would meet both its economic and creative goals.

The report lists next steps rather than new rules:

  • Gather more evidence on how copyright affects AI development.
  • Leave the licensing market alone for now, monitor it, and improve access to datasets, including through a planned Creative Content Exchange.
  • Work with industry on best practice for disclosing what AI models are trained on, which could later inform legislation.
  • Propose removing copyright protection for works created solely by AI, while keeping protection for AI-assisted works. This is a proposal, not a decision.
  • Explore protection for people’s voice and likeness, possibly through a new digital replica right.

No dates were attached to any of these steps.

What UK law says today

Today, UK copyright law has no exception for commercial text and data mining. Section 29A of the Copyright, Designs and Patents Act 1988 allows someone with lawful access to a work to copy it for computational analysis only “for the sole purpose of research for a non-commercial purpose”, and any contract term that blocks that use is unenforceable. Anything beyond that falls under ordinary copyright rules.

Pressure for a disclosure duty continued after March. A UK Parliament petition backed by the Creators’ Rights Alliance and joined by the Society of Authors asked the government to require AI developers to disclose the copyrighted works used to train generative AI. It closed on October 29, 2026. A petition is a request to government, not a change in the law.

UK and EU: what is law and what is only proposed (November 2026)
ItemStatus
UK: text and data mining allowed only for non-commercial research (CDPA s29A)Law
UK: opt-out exception for AI trainingDropped as preferred option, March 2026
UK: removing protection for purely AI-generated worksProposal
UK: mandatory training-data disclosureNot proposed; petition closed October 29, 2026
EU: AI model makers must respect opt-outs and publish training summariesLaw since August 2, 2025
EU: Commission fines for model makersPossible since August 2, 2026
EU: payment for training and an EUIPO opt-out listParliament recommendations, March 2026
EU: reform of the copyright rulesConsultations closed June 2026; no proposal yet

What the EU AI Act requires of AI model makers

The EU AI Act requires companies that make general-purpose AI models to follow EU copyright law and to publish a summary of their training content. Under Article 53, providers must have a policy to identify and comply with “a reservation of rights” made under the EU’s text and data mining rules, “including through state-of-the-art technologies”, and must publish “a sufficiently detailed summary about the content used for training” using a template from the EU AI Office.

The European Commission’s guidance sets the dates. These duties applied from August 2, 2025. The Commission’s enforcement powers, including fines, started on August 2, 2026. Models already on the market before August 2025 have until August 2, 2027.

The voluntary General-Purpose AI Code of Practice, finalized in July 2025, spells out what signatories promise on copyright:

  • Use web crawlers that follow robots.txt instructions.
  • Identify and comply with other machine-readable rights reservations that are widely adopted.
  • Not get around paywalls or subscription barriers.
  • Exclude websites that EU courts or authorities have found to be persistent commercial-scale pirates.
  • Offer rights holders a point of contact and a way to complain.

These rules sit on AI companies, not on authors. If you want to know what you must disclose when you use AI yourself, see our explainer on the EU AI Act and AI-assisted book labels.

The EU opt-out and the 2026 reform debate

In the EU, AI training on books you publish is legal under the text and data mining exception unless you opt out. Article 4 of the 2019 Copyright in the Digital Single Market Directive allows commercial mining only where the use “has not been expressly reserved by their rightholders in an appropriate manner”, and it gives machine-readable notices as the example for content published online.

Many authors see that opt-out model as too weak, and the debate moved forward in 2026. On March 10, 2026, the European Parliament adopted recommendations by 460 votes to 71, with 88 abstentions. MEPs asked for fair payment when generative AI uses copyrighted work, a full list of works used in training, and an opt-out list that the EUIPO could manage. They also want the Commission to examine how past use could be compensated. These are recommendations to the Commission. They do not change the law.

The Commission has started work on a possible reform. It ran a survey on how the 2019 directive has worked, open until June 14, 2026, and a call for evidence, open until June 25, 2026, on a future law provisionally titled a “Directive enhancing the copyright environment for European creativity and innovation”, according to the Kluwer Copyright Blog. The text and data mining rules are part of the review. No legislative proposal had been published as of this writing.

What UK and EU authors can do now

You can take several practical steps today, without waiting for new laws.

  • Add a rights reservation notice. Put a clear statement on your copyright page and on your author website that you reserve the use of your work for text and data mining and AI training. In the EU, this is how you use the Article 4 opt-out. Our free copyright page generator can draft the page around it.
  • Make it machine-readable where you can. On your own website, use robots.txt to disallow AI crawlers. The EU Code of Practice commits signatory AI companies to follow robots.txt and other widely adopted machine-readable reservations.
  • Keep your records. Save dated manuscript files, publication dates and ISBNs. If you ever need to prove ownership or show when you reserved your rights, records help. Our guide on how to copyright a book explains what protection you already have.
  • Register with ALCS if you are a UK writer. The Authors’ Licensing and Collecting Society collects money for secondary uses of writers’ work. It is also a development partner, with publishers’ body PLS, in a collective generative AI training licence that the Copyright Licensing Agency announced in April 2025. An opt-in collective licence would pay authors when AI developers license their work.
  • License only on terms you choose. If an AI company or a publisher offers a licensing deal, read what it covers: training, fine-tuning, and whether the deal can be ended. You do not have to accept.
  • Consider a human-authorship label. Society of Authors members can sign a declaration and use the “Human Authored” logo. Spelling and grammar checks, brainstorming and research help are allowed. The scheme is currently limited to members and text-based works.

A rights reservation tells AI companies that follow EU rules not to mine your work. It cannot remove books that were copied before you added it, and it does not stop companies that ignore the rules.

What to watch in 2027

The next big moves are a possible EU copyright proposal and any UK follow-up to the March report. In the UK, the government has promised more evidence gathering and work on transparency, but it has set no timetable. In the EU, observers expect the first clear signals from the Commission late in 2026 or in 2027, and the AI Act’s copyright duties now carry the threat of fines.

For indie authors, the practical position is stable: your copyright is intact in both the UK and the EU, an opt-out is your main tool in the EU, and any payment for AI training is likely to come through licensing, either direct or collective. If you write with AI tools yourself, for example with CruxOne’s AI book writer, the same rules protect the parts you create and revise: AI-assisted work stays protected in the UK, while the government has only proposed removing protection for works generated entirely by AI.

An author’s desk with a printed copyright page, a laptop, a pen and a cup of tea
A clear rights reservation on your copyright page and website is one of the few steps every author can take today.

Frequently asked questions

Is it legal in the UK for AI companies to train on my book?

UK law only allows text and data mining without permission for non-commercial research, under section 29A of the Copyright, Designs and Patents Act. The government dropped its plan for a broader opt-out exception in March 2026, and the law has not changed since.

Did the UK pass a new AI copyright law in 2026?

No. The March 18, 2026 report announced more evidence gathering, work on transparency and some proposals, such as removing protection for works created solely by AI, but no new legislation.

How do I opt out of AI training in the EU?

Reserve your rights “in an appropriate manner”, as Article 4 of the 2019 directive puts it. For content published online, a machine-readable notice is the stated example. In practice, add a clear statement to your copyright page and website and block AI crawlers in your site’s robots.txt.

Does the European Parliament vote mean AI firms must pay authors?

Not yet. The March 10, 2026 vote adopted recommendations to the European Commission. Payment would need new legislation, and the Commission has not proposed any so far.

Can ALCS get me paid for AI training?

ALCS collects secondary-use income for writers and is working with the Copyright Licensing Agency and PLS on a collective generative AI training licence announced in April 2025. Check ALCS directly for whether it is open and how payments work.

Sources

  1. UK government: Report on copyright and artificial intelligence (March 18, 2026) — checked 19 November 2026
  2. Copyright, Designs and Patents Act 1988, section 29A — checked 19 November 2026
  3. Euronews: Don’t Steal This Book, authors protest AI at London Book Fair — checked 19 November 2026
  4. downthetubes.net: Don’t Steal This Book, AI protest at London Book Fair — checked 19 November 2026
  5. downthetubes.net: Petition calls for transparency on AI training data — checked 19 November 2026
  6. Copyright Licensing Agency: Development of a Generative AI Training Licence with ALCS and PLS — checked 19 November 2026
  7. European Commission: Guidelines for providers of general-purpose AI models — checked 19 November 2026
  8. General-Purpose AI Code of Practice: Copyright chapter — checked 19 November 2026
  9. European Parliament: Protecting copyrighted work in the age of AI (March 10, 2026) — checked 19 November 2026
  10. Kluwer Copyright Blog: Europe’s next copyright reform and the Commission consultations — checked 19 November 2026

This article explains the rules in plain language. It is not legal advice. For decisions about licensing or enforcing your rights, talk to a lawyer or your authors’ organization.

CruxOne Editorial Indie publishing desk

We write about self-publishing for authors in North America and Europe: Amazon KDP, AI rules, royalties and the business of books. Every figure is checked against its source, and each source is listed on the page.

Related reading

Write, design and publish your book in one place

CruxOne helps indie authors write with AI, make the cover, format the interior and publish to Amazon KDP. Start free.