A man in Zhuji, Zhejiang, China, was rushed to intensive care after eating wild mushrooms he had collected near his home, despite an AI tool reportedly identifying them as non-toxic.
According to Zhuji Daily, the man had picked the mushrooms himself and used AI to check whether they were poisonous. After receiving what he believed was confirmation that they were safe, he cooked and ate them for lunch.
However, several hours later, he began experiencing stomach discomfort followed by five episodes of watery diarrhoea. He also developed severe soreness in both arms, prompting him to seek medical attention.
Medical examinations found that his creatine kinase level had reached 28,771 U/L, reportedly more than 100 times the normal level. Doctors also detected acidosis and elevated lactic acid levels, indicating serious complications including acute kidney injury and rhabdomyolysis.
After discussions with his family, the patient was urgently transferred to the ICU for treatment. Doctors focused on removing toxins, managing his symptoms and protecting his organ function. He eventually recovered and was discharged from hospital.
The case has also renewed concerns over relying on AI to determine whether wild mushrooms are safe to eat. In June, Douyin Group vice-president Li Liang had previously responded to a separate incident involving claims that an AI tool had mistakenly identified a mushroom as non-toxic.
On August 18, reporters reportedly tested several mainstream AI systems by asking them to identify poisonous mushrooms. The AI tools reportedly warned that they could not reliably distinguish toxic mushrooms from edible varieties and should not be used as the basis for deciding whether a mushroom is safe to consume.
The incident serves as a reminder that wild mushrooms can be difficult to identify accurately, and an AI-generated answer should never be treated as a guarantee of safety.
