AI news ·
Granta ends publishing partnership with Commonwealth prize after AI controversy
Granta ended its partnership with the Commonwealth short story prize after a regional winner faced accusations of AI-generated text. The magazine will no longer publish winning entries from the competition, citing a lack of editorial control.
Literary magazine Granta has ended its practice of publishing winning entries from the annual Commonwealth short story prize, a direct response to this year's controversy in which a regional winner faced widespread accusations of using AI-generated text. The magazine said it would stop all "external publishing partnerships" where it lacks editorial control, a decision that severs its long-standing relationship with the Commonwealth Foundation's flagship competition.
The 2026 prize cycle was thrown into turmoil in mid-May when critics on X and Bluesky flagged what they called "obvious markers" of AI writing in the Caribbean region's winning entry, The Serpent in the Grove by Jamir Nazir. The story drew scrutiny for its use of items arranged in threes, "not x, but y" constructions, and phrases such as "Sun on galvanise is a cruel instrument" and "She had the kind of walking that made benches become men."
In a statement to the Guardian, Granta said: "The 2026 selection of the regional winners of the Commonwealth prize caused a great deal of controversy, based on the speculation that one or more of the stories may have been at least partially AI-generated, accusations that were strongly rejected by the authors." The magazine added that it will keep the shortlisted stories on its website "in the public interest" but confirmed its editorial partnership with the foundation is over.
The author's defense
Nazir rejected the AI allegations, describing a writing process constrained by chronic health conditions. "My writing process is unusual," Nazir told the Observer via email. "It is conducted entirely on an Android phone. This is a necessity driven by chronic health conditions which make sustained, desk-bound typing physically impossible. That is why I rely on speech-to-text to do my writing, followed by minimal keyboard editing, along with the same process of speech-to-text."
Nazir's explanation points to a Speech-To-Text workflow as the source of the stylistic patterns that critics flagged as AI-generated. He said he uses the same method in his professional life and applied it to produce his competition entry. The defense did not quiet the debate, however, as scrutiny of AI detection in literary submissions continues to intensify across the industry.
Institutional responses and prize money
Granta publisher and philanthropist Sigrid Rausing addressed the situation on 19 May with a statement that left the door open to ambiguity: "It may be that the judges have now awarded a prize to an instance of AI plagiarism - we don't yet know, and perhaps we never will know." On the same day, Commonwealth Foundation director general Razmi Farook said all shortlisted writers had personally stated no AI was used and the foundation had confirmed this upon further consultation.
The prize awards £5,000 to an overall winner and £2,500 to regional winners. According to the Sigrid Rausing Trust website, the trust provided £30,000 to the Commonwealth short story prize between 2014 and 2016. The Commonwealth prize did not respond to a request for comment on Granta's withdrawal.
Why this matters for writers
For working writers, the Granta decision signals a structural shift in how literary gatekeepers handle AI uncertainty. Rather than adjudicating individual cases, a major publication has withdrawn from a partnership entirely - citing editorial integrity as the reason. This creates a precedent: when a publisher cannot verify a manuscript's provenance, it may choose to exit the arrangement rather than absorb the reputational risk.
The controversy also exposes a practical problem that no detection tool has solved. A writer using AI for Writers tools or speech-to-text software can produce prose that reads as synthetic to experienced readers, even if no generative AI was involved. The stylistic tells that critics identified - repetitive structures, specific syntactic patterns - can emerge from dictation workflows just as easily as from large language models. For writers submitting to competitions, the burden of proof is shifting in ways that existing guidelines have not yet addressed.