"Don't Steal This Book" is 88 pages of author names followed by nothing. That blankness is the point.
Nearly 10,000 writers - including Nobel laureate Kazuo Ishiguro, bestselling crime novelist Richard Osman, graphic novel legend Alan Moore, and children's author Jacqueline Wilson - published the symbolic empty book at the London Book Fair this week. The pages after the names are blank, representing what the authors say will happen to new literature if AI companies keep training on copyrighted books without paying for them.
The protest was organized by Ed Newton-Rex, a composer turned copyright campaigner who has been one of the loudest voices arguing that the generative AI industry is "built on stolen work, taken without permission or payment." He is not wrong that AI companies have been vague about their training data. Most major model providers have acknowledged using books, articles, and web content in training sets without individual licensing agreements.
The UK Policy Fight Behind the Protest
The timing is deliberate. The UK government is currently evaluating a proposed change to copyright law that would let AI companies use copyrighted material for training unless the creator explicitly opts out. That is an important distinction: opt-out means your work is fair game by default. You have to actively find and use whatever mechanism each AI company provides to exclude your writing. For individual authors, many of whom do not have legal teams or technical staff, that burden is significant.
UK ministers face a deadline of March 18, 2026 to publish both a comprehensive economic impact assessment and a progress report on the copyright-AI consultation. The authors want the government to reject the opt-out model entirely and instead require AI companies to license copyrighted material the same way any other commercial user would.
Why Authors Are Angrier Than Musicians or Visual Artists
Books are particularly valuable to large language models (the AI systems behind tools like ChatGPT and Claude). A single novel contains hundreds of thousands of words of structured, high-quality prose - exactly the kind of data that makes language models better at writing. Music and images matter too, but text is the core fuel for the models that most people interact with daily.
The economics are also more fragile. Most authors do not earn substantial royalties. The median income for a professional author in the UK was around £7,000 per year according to recent surveys from the Authors' Licensing and Collecting Society. If AI-generated text displaces even a fraction of book sales, the financial impact on working writers is real and immediate.
What This Means for AI Tool Users
If the UK adopts a strict licensing requirement, it could set a precedent that spreads to the EU and potentially the US, where similar copyright battles are playing out in courts. That would raise the cost of training AI models, which could eventually mean higher subscription prices for tools like ChatGPT, Claude, and Gemini. It could also mean models trained on smaller, fully licensed datasets - potentially less capable but on firmer legal ground.
The blank pages of "Don't Steal This Book" are a stunt, but the policy question underneath is not. How governments resolve the tension between copyright holders and AI companies will shape what these tools can do and what they cost for years to come.