AI Controversy Ends in Rupture Between Boréal and Goncourt Candidate Thélyson Orélien
Publisher Les Éditions du Boréal has gone from defending Thélyson Orélien to ending its collaboration with him, shortly after the novel “C'était ça ou mourir” was withdrawn from the Prix Goncourt list. But the extent of the AI use remains unknown — and the author denies everything.
The Canadian publisher Les Éditions du Boréal announced on Wednesday 21 October 2026 that it is ending its professional collaboration with the 38-year-old author Thélyson Orélien. The reason is the conclusion the publisher itself now draws: “New information, confirmed over the past few days, indicates that artificial intelligence was used in the writing process” behind the novel “C'était ça ou mourir,” according to a statement reported by AFP via The Guardian.
The rupture comes barely a month after the accusations first became known in September, and only a short time after the publisher stood by the author. It is a concrete turning point in a case that has shaken the French-Canadian book industry — but also a case that shows how thin the documented evidence base still is.
How the Review Was Allegedly Conducted
According to Boréal's own account, the publisher gathered all working versions of the manuscript and engaged an external party to analyse the drafts. “The results obtained this week revealed traces of AI tags in the metadata of certain documents,” the publisher stated, as reported by AFP via The Guardian.
It is worth noting what this claim involves — and what it does not. It is the publisher itself that engaged the analyst, the analyst is not named, and the method has not been made public. None of the available sources provide independent confirmation that the metadata findings exist as Boréal describes them. What can be established is that the publisher is publicly withdrawing its earlier support and justifying this with the review.
At the same time, Boréal acknowledges the limits of its own knowledge. The publisher says it does not know the extent of the AI use in the writing process, and that “we will probably never get a definitive answer.”
The Context: The Goncourt Withdrawal and the Pangram Result
The case exploded when the novel was withdrawn from the shortlist for France's major literary prize, the Prix Goncourt, after tests with the AI detector Pangram reportedly showed a 100 percent probability that the text contained machine-generated content. This result, too, was reported through news agencies, not independently verified, and the method behind the Pangram test is not documented in the available sources.
When the accusations first became known in September, Orélien pushed back by claiming that racism lay behind the attacks on him. The publisher at that time sided with the 38-year-old author — right up until last week, when Boréal was still standing by him.
The Publisher's Reversal
It was thus the reverse of the original case that triggered the rupture. According to Boréal's account, the publisher initially attempted to work with Orélien to gather evidence of his innocence. The review of the manuscript versions was part of this effort. It ended without a conclusion — and according to the publisher, Orélien then announced that he would not provide further information.
The sequence matters: the publisher did not come out with accusations, but withdrew from an author it had itself defended, following a review it had itself commissioned — and with a finding it itself describes as unclear in extent.
The Author's Defence
Orélien still denies everything. In a statement to AFP, he called the accusations “slander and heresy,” and admitted to limited use: “Here and there I did use tools for research, particularly for foreign idioms,” he said according to AFP.
Speaking to the French newspaper Libération on Wednesday, he maintained his position: he said he used AI for historical research in the creative process. And he advanced a chronological defence: according to his own account to the newspaper, the book was begun in late 2017, and he completed a first draft of more than 300 pages in 2019 — several years before AI services such as ChatGPT existed. If the chronology is correct, the earliest drafts could not possibly have been written with today's generative tools. But the argument addresses only the beginning of the writing process, not the period leading up to publication, and neither the publisher nor the available sources can confirm the date.
In a separate matter — plagiarism accusations — Orélien has admitted to having committed “unacceptable mistakes” earlier in his career, without specifying what they consisted of. The details of the plagiarism accusations are not documented in the available coverage.
What We Do Not Know
There is much this case cannot answer:
- The extent of the AI use. Boréal itself says that “we will probably never get a definitive answer.” Orélien denies that any sentence is machine-generated.
- The analyst's method. The external party's name and analysis method have not been made public. The claim of “AI tags in the metadata” rests solely on Boréal's word.
- The Pangram result. The 100 percent probability was reported, not independently confirmed, and there is no public documentation of the test.
- The reliability of AI detection in general. None of the sources address how reliable such tools are — a well-known problem area for the entire industry.
- The plagiarism accusations. What “unacceptable mistakes” covers is unspecified.
- Orélien's racism claim. He claimed in September that racism lay behind the attacks, but the available sources do not go deeper into this.
Why the Case Matters
Beyond the fate of an individual author, the case illustrates a structural problem for the publishing industry: a publisher can break with an author on the basis of its own statements about metadata findings and a commercial detector's output — without any of this being independently verifiable. At the same time, it is the publisher itself that acknowledges it will probably never get a definitive answer.
The book industry thus has neither a robust tool for proving AI use, nor a process for exonerating anyone. Boréal chose to act on incomplete information; Orélien refuses to contribute more. The result is a rupture that both parties present differently — and which readers, for now, can only choose between two opposing narratives.