Two neuroscience papers published this year allegedly generated by AI misused real scientists' open brain-imaging data. The original researchers caught them — here are the 5 red flags that gave the fakes away.
· Reported by The Scientist and Retraction Watch · Assessments below are the researchers' professional judgments — alleged, not proven
AI generated papers just had their most embarrassing exposure yet. In August 2026, Oxford emeritus professor Dorothy Bishop was peer-reviewing a manuscript for a small neuroscience journal when she recognized something from the abstract alone: the dataset was hers. Raw fMRI brain scans from 50 volunteers doing language tasks — thinking of words, matching words to pictures — which she had published on the open repository OpenNeuro four years earlier. What followed unraveled into two published papers, one retraction request, and an ongoing Elsevier investigation.
What is an AI-generated fake paper? It is a scholarly manuscript that looks like genuine research — abstract, methods, figures, references — but was substantially produced by an AI system rather than by scientists who actually ran the experiments or analyzed the data. In this case, Bishop's assessment was blunt: the papers look convincing on the surface but contain fundamental errors suggesting the authors analyzed the data "beyond perhaps downloading a little bit of it to produce a figure" — if they analyzed it at all.
— Bishop reviews a manuscript for a small neuroscience journal, recognizes her own OpenNeuro dataset from the abstract, finds the dataset description doesn't match what her team actually did and that references don't support their claims. The editor is also suspicious. The paper is rejected.
— Expecting the authors to try elsewhere, Bishop searches their names and finds two papers from this year in the Journal of Neurolinguistics (an Elsevier journal), both purporting to use OpenNeuro brain-imaging data.
— The first paper, coauthored by Afzal Khan and Dawood Khan (affiliated with King Saud University in Saudi Arabia and Iqra University in Pakistan), purportedly used brain scans from a dataset created by Kathy Rastle, a University of London psychologist. Rastle concluded they probably never used her data at all.
— The second paper, by Afzal Khan alone, claimed to analyze four OpenNeuro datasets. Bishop raised detailed concerns on PubPeer; she described Khan's attempt at a response as "a total fiasco."
— Bishop asks the journal to retract the articles. Elsevier confirms it is investigating but says it "may take considerable time." The journal's editor, Nai Ding, has submitted the findings of the journal's own inquiry to Elsevier.
Note on fairness: Khan told Retraction Watch he was aware of the concerns, was "taking them seriously," and was "engaging directly and in good faith with the journal," declining specifics while the review is ongoing. He holds a PhD in linguistics and works in an English Language Skills department; his pre-2026 publications were in linguistics and English teaching, not brain imaging.
These are the concrete tells Bishop and Rastle documented — each one a check any reader can run:
There was also a sixth, social tell: the May paper's acknowledgements thanked Rastle and her colleagues for sharing their data — data Rastle believes was never used. She called it an attempt "to create a veneer of respectability."
Bishop's own advice, from her blog: check the data, the code, and the references; test the plausibility of the research; ask whether the article actually makes sense. Her honest caveat: "the frauds are getting so good that these methods now seem inadequate." Use this checklist:
This is the perverse twist: open science created the vulnerability. OpenNeuro exists so researchers can reuse expensive brain scans instead of re-collecting them. But public datasets are also perfect props for fabricated papers — the data is real, downloadable, and citable, which lends the fake paper a borrowed credibility. As Bishop put it, the case is "very upsetting for those of us who are into open data." Rastle pointed to the volunteers: "Participants allow their data to be shared openly to further scientific progress. They don't allow it to be shared openly to perpetuate fraud." If misuse continues, the rational response from dataset creators is to stop sharing — and the whole scientific commons shrinks.
Bishop has asked the Journal of Neurolinguistics to retract the articles. Elsevier told her it would investigate but that it "may take considerable time" — a pace Rastle criticized: "When a paper is just so obviously deeply flawed, and probably a fake paper, it really shouldn't take months to get it out of the scientific literature. It's just polluting the scientific record and creating confusion amongst readers and other scientists." For now, both papers remain in the published literature with an investigation pending — which is exactly why knowing the red flags matters.
1. What first made Dorothy Bishop suspicious of the manuscript she reviewed?
2. Which claim in the May paper was impossible?
3. In the August paper, one graph showed a participant using a second language for over 50 years. What was the dataset's actual maximum?
4. What did Afzal Khan say when Retraction Watch contacted him?
5. What did Bishop find when she examined the manuscript's analysis code?
No. Dorothy Bishop and Kathy Rastle believe the papers were AI-generated based on their professional judgment as the scientists whose datasets were allegedly used. Elsevier says it is investigating the Journal of Neurolinguistics papers, and the author says he is engaging with the journal in good faith. Treat the AI-generation claim as alleged until the investigation concludes.
OpenNeuro is a public repository where researchers share brain-imaging data openly so other scientists can reuse it. The datasets at the center of this case included raw fMRI scans from volunteers doing language tasks.
Dorothy Bishop is an emeritus professor of developmental neuropsychology at the University of Oxford. She spotted the misuse because she recognized her own dataset from a manuscript abstract while peer-reviewing it in August 2026.
Afzal Khan told Retraction Watch he was aware of the concerns, was taking them seriously, and was engaging directly and in good faith with the journal. He said he could not comment on specifics while a review was ongoing.
Open data depends on trust: volunteers share brain scans to advance science, not to become props in fabricated papers. If open datasets become raw material for AI-generated fraud, researchers may stop sharing — and everyone loses.
Sources: Retraction Watch, "When open data are misused" (October 1, 2026); The Scientist, "Alleged AI-Generated Papers Misuse Open Neuroimaging Data" (October 15, 2026); Dorothy Bishop's blog (deevybee.blogspot.com). Assessments attributed to the named researchers are their professional judgments; the Elsevier investigation is ongoing.
Ingenious AI publishes 21 live security platforms, and 100% of them run on permanent getingeniousai.com domains.