The Zimbabwean writer Simbarashe Kundizeza has denied using AI to write a prize-winning short story after an article in the Atlantic said software had flagged it as almost completely AI-generated.<\/p>\n
Kundizeza\u2019s story The Blue Fighter<\/a> was announced as the winner of the prestigious Caine prize for African writers on 6 October. Five days later, The Atlantic published an article<\/a> by the author and journalist Vauhini Vara titled: \u201cYet Another Literary prize winner Seems to Be AI.\u201d<\/p>\n Vara said that she had run the story through the AI-detection software Pangram, which itself uses AI, and it had deemed the story to be almost completely artificially generated. The exception, she said, was an epigraph from what is described as an unpublished poem by the South African writer Vangile Gantsho.<\/p>\n Vara said that when she read the story, she \u201cwas struck by the proliferation of language that is common in AI-produced writing: the not x, but y sentence constructions; the glut of metaphors; the platitudinous aphorisms\u201d.<\/p>\n She said: \u201cTaste is subjective, but trust me, this story is bad \u2013 maudlin and full of tired tropes.\u201d<\/p>\n Kundizeza responded in a statement on X<\/a>: \u201cFor the avoidance of doubt, I wrote \u2018The Blue Fighter\u2019. I did not use generative AI to write it. It is entirely my own work.<\/p>\n \u201cI look forward to continuing to work with the Caine prize<\/a> team, and to celebrating this recognition with the people who helped bring this story into the world. Most importantly, I will continue to write.\u201d<\/p>\n Journalist Vauhini Vara said she had run Kundizeza\u2019s story through the AI-detection software Pangram, which itself uses AI, and it had deemed the story to be almost entirely AI generated. Photograph: Andrew Altschul<\/p>\n Similar claims have been made recently about other literary prize winners. In July, the Trinidadian writer Jamir Nazir won the Commonwealth Foundation\u2019s short story prize<\/a> with The Serpent in the Grove, despite critics on social media claiming it had \u201cobvious markers\u201d of AI writing.<\/p>\n Last month, the Canadian-Haitian writer Th\u00e9lyson Or\u00e9lien\u2019s novel C\u2019\u00e9tait \u00c7a Ou Mourir (It Was That or Die) was removed from the longlist<\/a> of France\u2019s most prestigious literary prize, the Goncourt, after an anonymous online account alleged the book was \u201calmost entirely\u201d written by AI.<\/p>\n The Caine prize said in a statement<\/a> that it was reviewing the allegations that Kundizeza had used AI. It said: \u201cWhen such claims are made, it is our responsibility to look into them. We aim to do this by taking a human-centred approach, ensuring that writers are supported and cared for during the process.\u201d<\/p>\n Its chair questioned the reliability of AI-detection software itself.<\/p>\n Ellah Wakatama told the BBC\u2019s Front Row radio show: \u201cAs someone who chairs a prize, I\u2019m obviously not against AI, but what I am against is it being used as a final arbiter in this case, regardless of what the Atlantic says, because they have already proclaimed our winner to be a cheater without investigating it properly.<\/p>\n \u201cI am not happy for the work of writers, and in this case especially black writers, to form the testing ground for something that is at best still rough and ready.\u201d<\/p>\n Vara said in her article, and repeated in an interview with Front Row, that she had read a 2021 story by Kundizeza that was tonally and stylistically different from The Blue Fighter.<\/p>\n In his statement, he responded: \u201cTo suggest that a writer cannot write in different styles or genres, or improve his writing substantially after five or more years fundamentally misunderstands the craft of writing.\u201d<\/p>\n Vara pointed to Pangram\u2019s own research that found that false positives were rare, including for non-native English speakers writing in English, as well as a peer-reviewed journal article<\/a> that found Pangram had \u201cstrong accuracy\u201d when assessing mock academic papers created fully, partly or with no AI. However, she acknowledged to the BBC that there was \u201crelatively little third-party research\u201d on the accuracy of AI detection software.<\/p>\n