The case began with an anonymous account on X, ‘Balance ton Claude’, which claims to have submitted excerpts from the novel to Pangram, a piece of software designed to detect texts generated by artificial intelligence. The result was reportedly high enough to substantiate the accusation. Grasset, for its part, firmly stands by its author. Orélien disputes the interpretation of his writing and points out, in particular, that certain stylistic devices for which he is criticised are part of a Haitian and Caribbean literary tradition.
But the dam has now burst.
For now that the method exists, why stop at Orélien? All one needs to do is buy a novel, copy a few passages, run them through several detection tools and publish the results on social media. Bestsellers, award-winning novels and authors whose style suddenly seems ‘too polished’, ‘too repetitive’ or ‘too perfect’ are likely to become prime targets.
This phenomenon is already occurring abroad. Authors have seen their publications or contracts called into question following suspicions regarding the use of artificial intelligence. The case of Mia Ballard, in particular, led to the cancellation of an American publication; she denied having used AI herself. The author Jerry Falade also denied similar accusations after a contract was cancelled.
We may therefore be on the verge of witnessing something both fascinating and rather worrying: a veritable hunt for artificial text.
For there is a huge problem. Detectors are not truth-telling machines. Even tools reputed to be highly effective acknowledge a margin of error. Researchers also highlight the methodological difficulties involved in this type of detection. A text flagged as likely to have been generated by AI does not, on its own, constitute irrefutable proof that an author had their book written by a machine.
And here lies the paradox: we are now asking artificial intelligence to determine whether a human is human enough.
This could become quite mind-boggling.
Take a writer who uses ChatGPT to look up historical information. Is it still their book? Of course it is. Another asks it to correct twenty mistakes in a chapter. Is that artificial writing? A third submits a sentence and asks for ten different versions before choosing one and then rewriting it. Where exactly does collaboration with the machine begin? At what percentage should the reader be warned? 1 per cent? 10 per cent? 40 per cent?
And why should the writer be subject to a standard of technological purity that we no longer demand in most other creative fields?
A photographer works with autofocus, image-editing software, digital noise reduction and retouching. A musician uses synthesisers, samples and software capable of correcting a voice. A filmmaker today constructs a film using dozens of digital tools. Yet we continue to speak of photography, music and cinema.
Literature, too, will probably have to define what belongs to the tool and what belongs to the author.
But the immediate danger lies elsewhere: that of constant suspicion. An anonymous accusation can now precede an investigation, and the result of a software programme can become media evidence even before its reliability and the conditions of the test have been properly examined. The Orélien case shows just how quickly this process can spiral out of control: a few posts on social media were enough to turn a technical inquiry into a national scandal.
The publishing world was clearly unprepared. Several publishers interviewed recently explained that they do not systematically use detection software to examine manuscripts; in many cases, the relationship between author and publisher is still largely based on trust. At the same time, Robert Laffont is already planning to publish a novel by Sébastien Bailly in 2027, in which the use of AI in the creative process will be explicitly stated.
This is probably the real revolution.
The question will soon no longer be: ‘Do writers use artificial intelligence?’
Of course some do.
The real question will be: what does writing still mean when the machine becomes just another tool?
And whilst publishers, literary juries and writers search for the answer, another profession is likely to experience spectacular growth: that of AI hunters.
Following on from Thélyson Orélien, they now have thousands of novels to test.
And perhaps just as many authors to accuse.
