Caine Prize Winner Clears His Name in AI Dispute

0
33
Caine Prize Winner Clears His Name in AI Dispute

The Caine Prize for African Writing announced on Wednesday 11 November that it found “no evidence that the story was produced using generative AI”. Zimbabwean author Simbarashe Kundizeza therefore keeps the ÂŁ10,000 prize he won on 6 October for The Blue Fighter.


The board considers the case closed, ending a month of doubt that began with a detector’s score.

A Prize Under Suspicion

Kundizeza received the prize at a London ceremony. The story first appeared in The Shallow Tales Review, whose editor submitted it months after publication, he noted.
On 11 October, a magazine article said the detection tool Pangram had flagged nearly the whole story as AI-generated, and the Caine Prize opened a review that day.

Kundizeza rejected the claim on 14 October and said the story draws on his family’s experience. He said he was inspired by a poem by Vangile Gantsho and combined it with fictionalised details from around 2002, when Zimbabwe’s economic troubles began.
Pangram had rated all 25 previous Caine Prize winners, and Kundizeza’s 2021 story The Whistleblower, as human-written. The journalist who raised the question said every other past winner she tested also came out “100% human”.

What the Review Found

Interim Director Susi Owusu led a review that included interviews with the author, the 2026 judges and the magazine. An independent digital forensic investigator analysed the drafts’ metadata.
Kundizeza produced four Word drafts, representing up to 27 hours of work, and the documents show the story developing between April and June 2025.

Board chair Ellah Wakatama said “the evidence is clear: The Blue Fighter was human-written”. She also said an AI detector is not a valid measure of AI involvement.
For 2027, publishers must submit an AI declaration for each story, shortlisted writers must make one too, and judges will receive guidance, with submissions opening on 23 November. Some African writers argued that such accusations mainly target non-white writers.

Why Detection Scores Mislead

A Stanford study published in 2023 tested seven AI detectors on 91 essays by non-native English speakers. At least one detector flagged 89 of the 91, and all seven agreed on 18, a study by Stanford researchers found.


Essays by US-born eighth-graders were handled almost perfectly, which suggests the tools can penalise writers who learned English formally.

Pangram is a different tool, and the study does not show that it makes the same errors. The case still shows that a probability score cannot settle authorship alone.
Drafts, version histories and interviews gave the Caine Prize stronger evidence, though gathering them took a month. Other prizes have faced similar disputes, including the Goncourt withdrawal of Thélyson Orélien, who denies the allegation.

Keep up with Daily Euro Times for more updates

Read also:

Finland Clean AI Pitch Collides With Forest Clearing

Does Britain Need a Ministry for Artificial Intelligence?

EU Cybersecurity Rules Create a New Kind of Cyber Slop

LEAVE A REPLY

Please enter your comment!
Please enter your name here