However, the initial accusation did not come from him. It originated from the X account ‘Balance ton Claude’, which had fewer than a hundred followers when it claimed that Orélien’s novel had been written almost entirely by artificial intelligence. To back up this claim, the account relied on Pangram, an automated detector of AI-generated text. The problem is a significant one: to date, the identity of the person behind this account has not been publicly established. Le Monde, which investigated the matter, reports that it was unable to determine who was behind it.
But an accusation made to 92 followers does not automatically become a national story. It needs to be amplified. Le Monde specifically notes that the message would not have gone viral to the same extent without the involvement of several public figures with a much larger audience, including Samuel Fitoussi – described as one of the first to have shared it – and Charles Sapin, deputy editor-in-chief of Valeurs actuelles. The original post went on to exceed three million views.
This is where the story becomes more interesting than the simple question: ‘Did Orélien use ChatGPT?’ It raises the issue of our new culture of suspicion. An anonymous accusation surfaces, a piece of software produces a seemingly scientific percentage, media personalities pick up on it and, a few hours later, it is up to the accused to prove their innocence.
However, the results from AI detectors do not constitute a definitive literary assessment. In this very case, tests carried out using different tools did not yield consistent results. RTL reports that two other detectors produced radically different conclusions depending on the passages examined. Turning these algorithmic probabilities into absolute proof therefore amounts to giving the machine precisely the power that some claim to be fighting against: the power to decide what is human and what is not.
Orélien claims to have begun *C’était ça ou mourir* in autumn 2017 and completed a first draft in 2019, i.e. before the public explosion of generative AI systems such as ChatGPT. Grasset has publicly reaffirmed its confidence in him and denounced the categorical claims levelled at the author. Nor does this prove that no AI tools were ever used subsequently in the editorial process: it simply serves as a reminder that there is a gap between algorithmic suspicion and proof – a gap that public debate sometimes seems in a hurry to bridge.
Samuel Fitoussi is perfectly free to criticise Orélien, his novel, his style, or even to find the evidence put forward convincing. Literary criticism thrives on disagreement. But his role in fuelling this controversy deserves to be viewed for what it is: that of a powerful media catalyst for an accusation made under a pseudonym. This does not, however, mean he should be identified as the owner of the anonymous account: there is currently no public evidence to support this.
And this is perhaps the most striking paradox of this controversy. We claim to defend the authenticity of literature whilst accepting that a real, identified author, publishing under his own name, can be placed in the dock by a source whose identity we do not even know, on the basis of an algorithm whose verdict is presented as that of a lie detector.
The real debate therefore extends far beyond Thélyson Orélien. If, in future, any text can be fed into a machine, declared to be ‘AI’ at 90 or 100 per cent, and its author then summoned to prove that they actually wrote it, we are in the process of inventing a strange presumption of literary guilt. And it could soon affect everyone.
