A major controversy has erupted within the French literary scene after Canadian-Haitian author Thélyson Orélien, celebrated as a standout talent of the new publishing season, faced accusations of relying on artificial intelligence to produce his debut novel. The 38-year-old writer recently secured the Prix du Roman Fnac, the inaugural major honour of the literary autumn, for his work titled C'était ça ou mourir (It was either that or die). Adding to his rising profile, he has also been named a finalist for two of the most distinguished accolades in French literature, the Prix Goncourt and the Prix Renaudot, both scheduled to be announced in the coming weeks.
Anonymous Group Levels Claims Using Detection Software
The controversy gathered momentum when an obscure group operating under the name Balance ton Claude (Denounce your Claude, named in reference to Anthropic's AI assistant) asserted on social media platform X that the book was almost entirely the output of machine intelligence. In a public statement, the group maintained that it had examined various excerpts from separate sections of the manuscript and obtained an identical evaluation indicating one hundred percent artificial origin across the board.
While the exact membership and background of the group remain unknown, they revealed that their assessment relied on a digital tool called Pangram. The group described this software as the most reliable detection system available, contending that its margin of error is virtually nonexistent. To support their claims, they argued that benchmark tests demonstrated Pangram delivers an erroneous verdict on machine-created writing only once in every 24,400 instances.
Orélien Rebuts Accusations and Cites Creative Roots
Orélien, who was born in Haiti and relocated to Montreal in 2010, moved quickly to refute the assertions. Speaking to the daily newspaper Libération, the author emphasised that his lifelong passion has been literature and artistic expression, remarking that he could see no reason why he would ever feel the urge to enlist a machine to write on his behalf. He confirmed that he is currently assembling a detailed dossier alongside his editor to provide definitive evidence of his authentic authorship.
The writer stressed that his style is deeply rooted in Haitian and Caribbean heritage, noting that the imagery, rhythmic structures, and lyrical repetitions in his prose reflect an authentic spoken tradition that belongs entirely to him. Orélien also pointed out to Libération that he completed the opening draft of the book back in 2019, long before commercial generative AI tools became widely accessible to the general public.
Publisher Backs Author as Debut Book Reaches Wide Readership
The publishing house Grasset issued a forceful defence of the author, stating that the accusations represent a deeply hurtful attack against a writer whose personal path is remarkable and whose creative work remains original and potent. The novel traces the journey of a schoolteacher compelled to flee brutal gang violence in Haiti, recounting the perilous adventures that eventually lead him to safety in Canada, where he finally regains the liberty to live without perpetual fear.
The book has achieved remarkable commercial success, described by its Canadian publisher as its fastest-selling release in six decades, and international rights have already been cleared for translation into roughly 20 languages. Born in 1988 in the northern Haitian city of Gonaives, Orélien initially gained attention in his homeland through poetic works before practicing journalism after settling in Canada.
Broader Scrutiny Over the Reliability of AI Detectors
The unfolding dispute mirrors a recent crisis surrounding Trinidadian writer Jamir Nazir, whose story The Serpent in the Grove saw its Commonwealth Short Story Prize temporarily rescinded in July due to automated AI allegations. That honour was later reinstated once the competition organizers concluded that the prose was unquestionably the author's own genuine work.
Technologists and researchers have consistently cautioned that AI detection platforms such as Pangram suffer from notable technical blind spots. Rather than possessing direct knowledge of who drafted a document, these utilities look for statistical regularities and probabilistic sentence patterns typical of language models, which can frequently overlap with standard human phrasing. Empirical studies have shown that automated detectors routinely produce false alarms on human writing, especially when evaluating texts created by authors using non-standard idioms or regional dialects like those found across the Caribbean. Furthermore, these tools often fail to flag text produced by artificial intelligence if the passages have undergone human rephrasing or revision, leaving the boundary between independent authorship and automated detection open to intense dispute.
