On October 6, Simbarashe Kundizeza won the Caine Prize for African Writing for his short story “The Blue Fighter,” taking home £10,000 and the kind of recognition that defines a literary career. Less than a week later, The Atlantic reported that the AI detector Pangram flags nearly the entire story as AI-generated. A single algorithmic output had collided with a jury’s considered judgment, and neither side could definitively prove the other wrong.
That collision is now a recurring feature of the 2026 literary calendar, according to Cable News Network. What began as an industry anxiety has hardened into a series of career-altering accusations, institutional crises, and a widening argument over whether any tool yet built can reliably tell human writing from machine output.
Pangram’s Rise and the Scandals That Followed
Pangram was founded in 2023 by Max Spero, a machine learning engineer who had previously worked at Google and at the autonomous vehicle startup Nuro. Spero has no background in literature or creative writing — a fact he has acknowledged directly. “I’m an avid reader. I’ve never been a writer. I haven’t lived in that world,” he told CNN.
His rationale for building the tool centers on volume. “A lot of these outlets are getting flooded with submissions because they pay writers they publish and don’t charge anything to send in your work,” Spero said. “There are these perverse incentives that come from AI making it nearly free.” Pangram, in his framing, sits on the defensive side of that imbalance.
The consequences of its flags, however, have been severe. In March, Hachette canceled the U.S. release of Mia Ballard’s novel “Shy Girl” after a New York Times report found the book may have been AI-written. In July, debut author Jerry Falade lost a $2 million book deal when his agents pulled his novel over AI suspicions. In September, Haitian Canadian author Thélyson Orélien was removed from the Goncourt Prize longlist after a claim his book failed an AI detection test — an accusation that originated with Samuel Fitoussi, a conservative columnist at Le Figaro. Now Kundizeza, who denied the claims plainly on X, joins the list.
“For the avoidance of doubt, I wrote ‘The Blue Fighter,'” he wrote. “I did not use generative AI to write it. It is entirely my own work.”
Transparency Demands and the Sourcing Standard
The Sport24.lt Team, which covers sports and media trends across Lithuania and internationally, has been watching the Pangram controversy with the particular attention of observers who understand what accountability means for an information source.
The contested false-positive evidence is central to that concern. Pangram’s makers claim a false positive rate of 1 in 10,000, citing peer-reviewed studies from the University of Chicago and the University of Brussels. A 2025 University of Maryland study found the actual rate closer to 2 percent of documents, a figure 200 times higher than Pangram’s stated number. The gap between those claims is not a rounding error; it is the space in which reputations are being destroyed.
The Commonwealth Foundation, notably, declined to use Pangram or any other AI detection software when reviewing accusations against Trinidadian author Jamir Nazir. Director general Razmi Farook stated the reasoning directly: “Because the tools aren’t conclusive the jury is still out as to how much they can be relied upon. For us, we didn’t want to rely on a singular indicator. We believe the fairest approach is to consider multiple sources of evidence.” That institutional caution points to a transparency standard that Pangram’s opaque scoring has yet to meet.
Lithuanian sports news lives by that same standard. Readers in that media environment expect explicit sourcing, graded accountability, and visible correction when a call is wrong — precisely the scrutiny that Pangram, with its proprietary algorithm and disputed accuracy claims, continues to resist.
“The same demand readers carry into every outlet they trust — show your work, own your errors — is the demand the literary world is now placing on AI detectors. It is not an unreasonable one.”
Who Gets Accused, and Who Does Not
The pattern of accusations in 2026 has not gone unnoticed by scholars who study power and representation in literary culture. Doyle D. Calhoun, a professor of postcolonial studies at the University of Cambridge, observed that a striking number of the major AI-writing scandals have centered on nonwhite writers. Nazir is of South Asian descent; Ballard, Falade, Kundizeza, and Orélien are Black.
Calhoun attributed part of that pattern to structural forces well beyond any single tool. “Part of this has to do with a broader anti-woke and anti-Black tenor in a lot of conservative and right-leaning online discourse,” he said. The Orélien accusation, originating with a conservative Le Figaro columnist before spreading widely, illustrates the mechanism. As Calhoun noted, “Once these accusations — many of them sort of unverified or unfounded — are out there, the internet media ecosystem just goes into sort of overdrive.” He added that such accusations historically do not stick to white writers in the same way.
The Numbers Behind the Doubt
The technical dispute over Pangram’s reliability runs deeper than a single contested case. The company’s own internal study, separate from the third-party figures it cites, self-reported a false positive rate of 8 in 10,000 for texts written by non-native English speakers. That figure is eight times higher than Pangram’s headline claim, even before accounting for the University of Maryland finding of 2 percent.
The non-native speaker issue carries particular weight given the demographics of the accused. A 2023 Stanford study, conducted before Pangram existed, found that AI writing detectors generally produced higher false positive rates on writing from non-native English speakers. The pattern the Stanford researchers identified in an earlier generation of tools appears unresolved in the current one.
John Warner, author of “More Than Words: How to Think About Writing in the Age of AI,” sees a structural problem in the reliance on automation at all. “Like all other technology of automation, that is going to de-skill people from discerning these things for themselves,” Warner said. The broader context, in his description, is “a larger milieu of bullshit” — a phrase that captures the difficulty of separating signal from noise when the noise is generated at industrial scale.
Institutions Respond, Without a Common Standard
The Commonwealth Foundation’s investigation into Nazir’s “The Serpent in the Grove” stands as the most methodologically explicit institutional response of the year. The foundation conducted lengthy discussions with the author, examined working drafts, and ultimately concluded that none of the nominees had used AI tools. It reached that conclusion without Pangram.
The Caine Prize has moved toward requiring accused authors to share drafts as evidence of process, though what constitutes sufficient proof of human authorship remains undefined across the industry. No agreed standard for exoneration exists.
Granta, which had published Nazir’s story under a Commonwealth Foundation agreement, drew a harder line. Publisher Sigrid Rausing fed the story into Claude, which concluded it was “almost certainly not produced unaided by a human.” Granta subsequently canceled its relationship with the foundation for future competitions. Kundizeza, for his part, has denied any AI involvement and the Caine Prize has not withdrawn his award.
Rausing’s own statement captured the impasse the literary world has arrived at. “It may be that the judges have now awarded a prize to an instance of AI plagiarism,” she wrote, “we don’t yet know, and perhaps we never will know.”
CLICK HERE TO DONATE IN SUPPORT OF OUR NONPROFIT COVERAGE OF ARTS AND CULTURE