My novel was stolen by a robot – and used to train AI without my consent

I’ve been robbed. Along with Zadie Smith, Stephen King, Maggie O’Farrell and thousands of other writers. We don’t know exactly when, and we couldn’t describe the thief. But sometime in the last year our books were snatched from the shelf by someone in Silicon Valley and forced into a gigantic algorithm, which is now being used to generate unauthorised content and unlimited profits for AI systems including ChatGPT.

The thief, or thieves, almost got away with it. But they left traces – ghosts in the machine, caught by hard-working humans. A piece in the Atlantic – now read furiously by almost every writer on the planet – revealed more than 191,000 books are being used without permission to develop AI systems by Meta, OpenAI and other companies. A whole stolen library. And these are just the ones we know about so far.

I did not give consent. I have been given no credit and offered no compensation. I still can’t quite believe this is happening. I first saw the story tweeted by Sathnam Sanghera who wrote: “So my books are also being used to train AI without my permission … Millions of hours of authors’ work being exploited by big tech with zero payment. Funny how writers keep getting shafted. RAGE”.

I checked the list and there I was. It was like walking into Waterstones and watching every shelf being emptied and nobody doing anything to stop it. Rubbernecking, I checked for my favourite writers. They’re all there. This might be the only time I’m on a list with Margaret Atwood and Maya Angelou. Maybe the tech companies are relying on the fragile egos of writers to get away with what may soon be decided by courts is an actual crime.

Stolen books include fiction and non-fiction. They range from crime to romance, from niche names to bestsellers, including winners of the Booker, Pulitzer and Women’s Prizes. Ian Rankin, Val McDermid and Denise Mina are all victims – that’s almost a whole genre. Even the dead aren’t safe – Jane Austen, Walter Scott and Ray Bradbury are all in there (although to be fair to Bradbury he probably saw this coming).

“Eighteen of my books are on this database,” tweeted Joanne Harris. “And have already been used to train AI without my consent.” Harris is chair of the management committee of the Society of Authors, the union for writers. Never have we needed unions more.

Our collected works have been mulched into a data set known as “Books3”, which sounds like a sci-fi baddy. Millions of carefully crafted sentences zapped in an instant into mere raw material which can now be used to build almost anything: a questionnaire to root out “fake” asylum seekers, a persuasive “end of life” chatbot, or simply acres of fakes which will further impoverish writers and eventually starve readers.

“Books3” includes my first novel, You Will Be Safe Here, which is set in South Africa in 1901 during the second Boer war, and the present day. It explores the British concentration camps there linking this still-hidden history to contemporary conversion therapy camps. But, break my novel down into data – turn it upside down and inside out and divorce it from my values – and my rejection of imperialism, racism and homophobia could easily become a “how to” guide. Not a light in the dark, but a deeper darkness. It is sickening and unethical – but not unavoidable.

I did not give consent – had somebody bothered to ask me or my agent I would have refused. I have not been paid – not one penny, not the merest whiff of royalties, not even the chance to win an Amazon voucher.

My name doesn’t appear on any of these banal products, even though they could not exist without all the years of research, writing and editing I have done. This is theft, pure and simple. But on an industrial scale.

Writers are not refusing progress – we’re not nostalgic for quills. We just want fairness and quality for writers and readers. Do you want to read a novel written by a computer? Watch a film “imagined” by something without an imagination? This is why the WGA went on strike in Hollywood for 146 days. And they won: AI can’t write or rewrite literary material. Writers there are safe (for now).

But what about here? Let’s not forget, AI can’t yet do anything on its own. Machines can only do what we tell them (again, for now). So we must look behind the curtain. Or instruct our lawyers to do so.