The French book scene is in uproar after a Canadian-Haitian author billed as a sensation was accused of using artificial intelligence (AI) to write his first novel. Thélyson Orélien, 38, has just been awarded the Prix du Roman Fnac – the first of the literary season's prizes – for C'était ça ou mourir (It was either that or die). He is also a finalist for the most prestigious annual French language awards, the Goncourt and the Renaudot, set to be issued soon.
But on Tuesday a little-known group called Balance ton Claude (Denounce your Claude, in reference to Anthropic's AI assistant) claimed the book was "almost totally" written by AI - a charge the author denied. "We tested several passages, from different bits of the manuscript, to come up with the identical result: 100 percent AI," the group said on X. According to the group – whose membership is unclear – the detection was carried out by a programme called Pangram, which it said was "by far the most accurate", with an almost negligible chance of making a mistake.
Orélien, who was born in Haiti and moved to Montreal in 2010, immediately denied the accusation. "I have always loved creation and literature. I can't see why I would feel compelled to use a machine," he told Libération newspaper.
He said he was "compiling a dossier" with his editor to prove that he authentically wrote the book. "My traditions are Haitian and Caribbean. The images, the repetitions and the rhythms are the living word, they are my own," he said.
The publishing house Grasset also came to his defence, saying the allegations "are atrociously injurious to a writer with an exceptional life-story" whose "work of creation is original and powerful." The novel tells the story of a teacher who flees gang violence in Haiti and his adventures before he finally makes it to Canada, where he re-discovers "the right to breathe without fear". The row recalls the controversy over Trinidadian writer Jamir Nazir's The Serpent in the Grove, whose Commonwealth Short Story prize was initially withdrawn in July over allegations that it was written with generative AI. The prize was subsequently re-awarded, after the organisers said they were satisfied the writing was genuine.
Critics of Pangram and other AI-detectors say they are unreliable – with different programmes often giving different results after analysing the same texts. These tools do not actually know who, or what, wrote a piece of text. Instead, they look for patterns commonly found in AI-generated writing, but those patterns can also appear in human work.
Extract — continue reading at the source.
