The literary world is at war. On one side, the writers and publishers devoted to the human craft of writing. On the other, those who wield machines that can spit out a novel in seconds. And now a French prize has raised the stakes.
The Prix Pascal de Littérature IA opened for entries a couple of weeks ago. It is designed to celebrate the best fiction written with AI. Stick a prompt into ChatGPT, and you could be rewarded with help getting your manuscript into print. The prize will be awarded to a work of fiction written by a large language model (LLM), and the human behind it must also submit the prompts and process they used—which, in a strange reversal, will prove that a machine, not a person, did the writing.
At a time when writers are organising against machine-made novels, this is a dispiriting provocation.
And it arrives in the wake of a number of AI-related scandals in the literary world. Popular authors, some of them prize-winners themselves, have been accused of passing off machine-made prose as their own.
Last month, the Canadian publisher Les Éditions du Boréal cut ties with the Haitian-Canadian author Thélyson Orélien, after he was accused of using AI to write C’était ça ou mourir (It Was That or Die) on the very same day he won the Prix du roman Fnac. Orélien denied using AI in his writing, although he said he used it for research and did admit plagiarism in earlier work.
Earlier this year, the pitchforks came out when Granta published “The Serpent in the Grove” by Jamir Nazir, the Caribbean regional winner of the Commonwealth Short Story Prize, which included such confusing lines as “She had the kind of walking that made benches become men”. After an investigation, the Commonwealth Foundation said it was satisfied that Nazir had not used AI and went on to name him the prize’s overall winner. Granta, for its part, decided to stop entering publishing partnerships over which it has no editorial control.
Beyond the immediate deceit of passing off machine writing as human, there is another problem. Authors accused of using AI, whether they have or not, will find the mark hard to shake off. They wear a modern scarlet letter.
The emergence of AI in literature is so insidious that the Society of Authors (SoA) launched a “Human Authored” logo scheme in March, so that authors can declare that their work is not written by a machine. Beyond the is-it-isn’t-it-AI novels, platforms such as Amazon are awash with books generated entirely by AI: novelty books, romances and self-help titles, among many others.
Confusingly, in mid-October, the SoA also helped launch an ethical AI charter with other international writers’ organisations, designed to set out “commitments to the responsible and ethical use of generative AI services”. Barely a week later, after a backlash from its own members, it had withdrawn.
“By signing up to the Charter, industry organisations such as agents, publishers and prize bodies can pledge to only use generative AI in a transparent, responsible and ethical way,” the SoA wrote earlier in October.
But when AI companies are training their models on stolen work, the word “ethical” rings hollow.
Earlier this year, the Don’t Steal This Book campaign, backed by the SoA among others, published a paperback containing 88 pages of the names of authors pledging their support, followed by 86 blank pages. Those blank pages stand for what will be left if the UK government legalises “book theft to benefit AI companies”, the campaign website says.
Generative-AI systems can only grow on what they are fed. Books are a staple of that diet—whether or not their authors have consented. An investigation by the Atlantic revealed that more than 191,000 pirated books had been used to train some of the biggest AI models, with the authors neither consenting nor receiving payment. The titles ranged from self-published erotica to bestselling novels, and some of the authors have brought lawsuits against Meta for copyright infringement.
It doesn’t end there. With that pesky detail of copyright in the way, the AI company Anthropic has been found to have taken a different approach to training its chatbot, Claude. As well as downloading pirated books (the subject of a $1.5bn settlement with authors last year), it spent tens of millions of dollars buying used books, cutting them up, scanning them in, and then destroying them. As if the literal destruction of books were not on the nose enough, the ex-books were then reportedly shipped off to be recycled into things like toilet paper.
As generative-AI systems gorge themselves on human writing, the livelihoods of real writers are coming under threat. A number of ghostwriters recently told the Bookseller that they have seen a sharp reduction in enquiries, as people use AI to write their books instead, and that people are also presenting them with AI-generated manuscripts to edit.
If there’s one thing LLMs don’t need, it’s encouragement. A prize celebrating AI writing sends entirely the wrong message about what is important in creative writing: the connection made between one human being and another. Not how human it’s possible for a machine to appear. Strangely, the Prix Pascal insists it is “not a technology competition”.
What we should do instead is celebrate works that interrogate the impact of AI on humanity. In Dead but Dreaming of Electric Sheep, Paul Tremblay confronts the scourge of AI-generated writing head-on, in interludes that punctuate a story about a comatose man whose body is piloted by AI for undisclosed purposes. Grace Chan’s Every Version of You explores a dark scenario in which reality has fused with a virtual world, and where what it means to be human grows ever less clear.
Instead of pouring time and money into prizes that threaten our creative industries, we should be strengthening the protections that allow human writing to flourish. When there are so many talented authors in the world, why waste time honouring something that rewards those who can’t write and threatens the craft itself? Literary awards are for writers, not machines.
