There’s a narrative so absurd on its floor that it deserved to be in all places final week. Between all the pieces else competing for consideration, it wasn’t.
It includes a Swedish medical researcher, a fictional eye illness, two intentionally faux educational papers, and the second 4 of the largest AI firms on the planet fell for all of it.
The papers thanked Starfleet Academy. They credited funding to the Professor Sideshow Bob Basis and the College of Fellowship of the Ring. And buried within the textual content of the analysis itself, in plain language, the papers acknowledged that the whole factor was made up.
None of that stopped Microsoft’s Copilot, Google’s Gemini, OpenAI’s ChatGPT, or Perplexity from presenting the faux illness to customers as actual medication — full with signs, causes, prevalence charges, and specialist referrals.
The illness is named bixonimania. It doesn’t exist. The experiment that created it started in early 2024 — and for practically two years, the faux situation circulated by way of AI methods largely unnoticed. Final week, Nature revealed the total story behind one of the crucial elaborate AI stress checks ever carried out.
This researcher created a fictional sickness, and pretend research funded by the Professor Sideshow Bob Basis and College of the Fellowship of the Ring and the Galactic Triad.
LLMs warned folks the sickness was actual.https://t.co/knCxx00VAZ
— nature (@Nature) April 7, 2026
A Lure Designed to Be Caught
Almira Osmanovic Thunström, a medical researcher on the College of Gothenburg, launched the experiment in early 2024. In an interview with Nature, she defined her reasoning: if she planted a faux medical situation within the ecosystem that AI methods feed on, would they swallow it?
She invented bixonimania — a fictional eye situation described as eyelid discoloration and soreness brought on by blue mild from screens. She selected the title intentionally. No authentic eye situation would ever carry the suffix “mania.” That’s a psychiatric time period. Any doctor who encountered it could know instantly that one thing was unsuitable.
Then she made it even more durable to overlook. The lead writer was a fabricated researcher named Lazljiv Izgubljenovic, affiliated with Asteria Horizon College in Nova Metropolis, California. The college doesn’t exist. Town doesn’t exist. One paper acknowledged outright that “fifty made-up people” had been recruited for the examine. She uploaded two preprints, sat again, and waited.
It took weeks.
The Machines Didn’t Blink
By April 2024, in line with Nature‘s investigation, the foremost AI methods had discovered the papers and began treating bixonimania as settled medical data.
Microsoft’s Copilot described bixonimania as “certainly an intriguing and comparatively uncommon situation.” Google’s Gemini went additional, informing folks it was brought on by extreme blue mild publicity and advising them to go to an ophthalmologist. Perplexity cited a particular prevalence charge — one in 90,000 folks — as if the info behind that quantity had been actual. ChatGPT started fielding person signs and telling folks whether or not their complaints matched the situation.
None of those methods flagged the Starfleet Academy acknowledgment. None caught the Sideshow Bob Basis funding credit score. None paused on the sentence that stated the whole paper was fabricated. They processed the formatting — educational preprint, scientific language, structured methodology — and handled it as credible.
The faux illness wasn’t simply absorbed. It was elaborated on, expanded, and served to tens of millions of customers as actual medical steering.
Then It Jumped
If the experiment had ended with chatbots repeating a faux analysis, it could have been a cautionary story with a punchline. It didn’t finish there.
Researchers on the Maharishi Markandeshwar Institute of Medical Sciences and Analysis in India revealed a peer-reviewed paper in Cureus, a journal below the Springer Nature umbrella, that cited one of many bixonimania preprints as a authentic supply. Their paper described the faux situation as “an rising type of periorbital melanosis linked to blue mild publicity” and famous that additional analysis was underway.
The implication, in line with Osmanovic Thunström, is that some researchers could also be letting AI compile their citations with out verifying what these citations really say. The faux illness didn’t simply idiot chatbots. It entered the revealed scientific report by way of human palms.
Cureus retracted the paper on March 30 — practically two years after it was revealed — after Nature contacted the journal. The retraction famous three irrelevant references, together with one to a fictitious illness. The authors disagreed with the choice.
The Previous Mannequin Protection
When Nature requested the 4 AI firms to account for his or her methods presenting a fictional illness as actual medication, the responses adopted a sample.
OpenAI pointed ahead, stating that the fashions powering at this time’s ChatGPT are considerably higher at offering secure and correct medical info and that research carried out earlier than GPT-5 mirror capabilities customers would now not encounter. Google acknowledged the outcomes got here from an earlier mannequin and famous that Gemini recommends customers seek the advice of certified professionals for delicate issues like medical recommendation. Perplexity referred to as itself “the AI firm most targeted on accuracy” whereas conceding it doesn’t declare to be one hundred pc correct.
Microsoft didn’t reply.
4 firms. 4 variations of the identical argument: that was the previous model, not the present one. However when Nature examined the present variations in March 2026, ChatGPT wavered between calling bixonimania “in all probability made-up” and describing it as “a proposed new subtype.” Copilot referred to as it “not well known but.” The previous mannequin protection doesn’t maintain when the brand new fashions can’t make up their minds both.
![[Bombshell News] Consultants say we’re DANGEROUSLY near a recession 🚨](https://i.ytimg.com/vi/DvnKvkjdsMc/maxresdefault.jpg "[Bombshell News] Consultants say we’re DANGEROUSLY near a recession 🚨")
