A celebrated novel by a Canadian-Haitian author has been removed from the longlist of France’s most prestigious literary award, the Prix Goncourt, following allegations that it was generated largely by artificial intelligence. The case involves an anonymous social media accusation, disputed detection software, a separate plagiarism claim, and a writer who denies everything. What it reveals goes well beyond one book or one prize.
How an Anonymous Post Upended a Literary Career
Thélyson Orélien’s novel C’était ça ou mourir (translated as It Was That or Die) had already achieved considerable success before the controversy began. Published in France by Grasset in August, the book sold around 35,000 copies and won the Prix du roman Fnac. Translation rights had been sold to more than 20 countries. The novel tells the story of a history and geography teacher who flees gang violence in Haiti and travels through 12 countries before reaching Canada. It is written in the first person, in a lyrical register that Orélien, 38, describes as rooted in Haitian and Caribbean language traditions.
On 21 September, the same day the book won the Fnac prize, an anonymous account on X called “Balance ton Claude” (a reference to Anthropic’s AI model Claude) claimed to have run the text through an AI detection tool called Pangram and found it to be “almost entirely” written by artificial intelligence. The allegation spread quickly. The author was later identified as Samuel Fitoussi, a columnist for the French centre-right newspaper Le Figaro and the author of an essay against what he described as “wokeism” in film and television. On his own account, Fitoussi stated that his concern was not hostility to Orélien specifically, but a fear that AI would cause “the disappearance of reading and writing.”
Detection Tools, Plagiarism Claims, and a Shifting Rationale
Here is what most coverage of this story underplays: the AI detection allegation alone did not remove the book from contention. Philippe Claudel, president of the Académie Goncourt, had initially urged caution about the AI claims. French media outlets tested other detection programmes and found that several identified the novel as very probably written by a human. Questions about Pangram’s accuracy were raised publicly. The detection software, in other words, produced contradictory results depending on which tool was used.
What appears to have changed the academy’s position was a second, separate allegation: that Orélien had plagiarised a short story from the preface of The Morning of the Magicians, a book by French journalist Louis Pauwels originally published in 1960. This plagiarism claim was published by the newspaper Libération. The Académie Goncourt’s formal statement cited both issues together, referring to “plagiarisms of which its author has been guilty over time” alongside the conclusion, drawn from what it called “trustworthy researchers and journalists,” that the book is “in all likelihood very largely the product of artificial intelligence.”
Orélien has denied the AI allegation and described it as racially motivated. His Canadian publisher, Les Éditions du Boréal, stated that it could “neither confirm nor deny” the allegations at this stage, while condemning what it called “hateful messages and personal attacks.” Orélien, speaking at a literary festival in Manosque, said he perceived “a desire not only to attack the book, but especially my voice” and announced he would step back from public comment until the situation settled.
What This Case Exposes About AI, Authorship, and Institutional Pressure
The deeper issue here is not whether this particular novel was written by AI. It is that no reliable method currently exists to answer that question with certainty, and institutions are being forced to make consequential decisions in the absence of that certainty.
AI text detection is a genuinely unsolved problem. Tools trained to identify machine-generated text produce inconsistent results across different software, and they are known to flag certain writing styles, including non-native or stylistically distinctive prose, at higher rates than others. This is not a minor technical footnote. It means that an accusation, even an anonymous one, can now carry institutional weight that outpaces the evidence behind it.
The Académie Goncourt’s decision was framed as protecting the integrity of a prize whose purpose is to honour literature “written by women and men.” That framing is understandable. Literary prizes depend on the assumption of human authorship. But the mechanism that triggered the review was an anonymous social media post citing a single, contested detection tool. The plagiarism allegation added a separate and more concrete dimension, but the two issues became entangled in public discourse in ways that made them difficult to disentangle.
This matters beyond literature. In any field where AI-generated content is possible but not always detectable, the gap between accusation and proof becomes a space where reputations can be damaged before evidence is established. The tools for detection have not kept pace with the tools for generation.
In Short
An anonymous social media allegation, amplified by disputed AI detection software and compounded by a separate plagiarism claim, was enough to remove a commercially successful, critically recognised novel from one of the world’s most prestigious literary longlists. The case does not prove the book was AI-generated. It proves that the absence of reliable detection methods creates a vulnerability that institutions, prizes, and individual authors are not yet equipped to navigate.
Based on reporting from The Guardian - Technology.