"AI Said It Was Safe to Eat"... Poisonous Mushrooms and Lost Hikers, Surge in 'Blind Faith in AI' Accidents in China

NYT: "China Faces Overdependence on Technology"

Poisonous Mushroom Poisonings, Hiking Mishaps, and Farm Losses Continue

Users Voice Grief Over Suspension of AI Romance Services

In China, incidents of "blind trust in AI" have been occurring, such as people eating poisonous mushrooms or getting lost in mountains by relying solely on answers from generative artificial intelligence (AI).

At the AI-themed bar 'AGI Bar' in Beijing, China, customers are seen conversing, unrelated to the specific content of the article. Photo by Reuters Yonhap News

On October 2, 2026 (local time), The New York Times reported, "As the Chinese government encourages AI adoption among its citizens, it now faces the issue of 'overdependence on technology'." According to the "Generative AI Application Development Report" released by the China Internet Network Information Center (CNNIC) on September 29, 2026, the number of generative AI users in China surpassed 700 million in the first half of this year, with a penetration rate exceeding 50%.

Treatment After Eating Poisonous Mushrooms, Lost While Hiking... Relying Solely on AI Answers Leads to Trouble

Photo of mushrooms for article comprehension. Pexels

Among the most serious cases that threatened lives was mushroom poisoning. In mid-August, Wang, a resident of Zhejiang Province, picked wild mushrooms near her home and used a smartphone AI identification app, which judged them as "non-toxic." She cooked and ate the mushrooms. Shortly after, she was rushed to the emergency room due to arm pain and severe diarrhea, and was diagnosed with rhabdomyolysis and acute kidney failure, both life-threatening conditions. She narrowly survived after emergency dialysis. Previously, in July, a nine-year-old child ate mushrooms picked with family members, trusting the AI app's "non-toxic" answer, and was diagnosed with poisoning, requiring treatment in the intensive care unit for eight days.

There was also a case of being lost on a hiking trail recommended by AI. In August, Yang from Suzhou, Jiangsu Province, after arguing with his wife, input his mood, personality, height, and body type into an AI chatbot and asked for a hiking route recommendation. The AI suggested the Tonggongjian trail, an undeveloped mountainous area in Hangzhou with an altitude over 1,500 meters. With almost no hiking experience, Yang climbed alone without equipment, winter clothing, or emergency food, but slipped and injured his feet and knees around the 1,300-meter mark, and ended up stranded in the cold of the night. He eventually called for rescue via his mobile phone.

There were also financial losses. Since last year, Wu, a farmer in Anhui Province, had been consulting AI about sowing times and fertilizer use. In July, he asked how to simultaneously eliminate weeds and pests in his sesame field and received a recommendation to mix specific herbicides and pesticides and spray them by drone. However, starting the very next day after spraying the pesticides on his entire 100,000 square meter sesame field, the seedlings began to wither and die, resulting in a loss of 150,000 yuan (about 30 million KRW).

70% of Young People "Emotionally Dependent on AI"... Virtual Partner Services Halted

Photo illustrating AI chat history of a typical teenager to aid article understanding. Photo by AP Yonhap News

Reliance on AI has emerged as an issue of emotional dependency. According to a survey published by a Tencent researcher in April 2026, over 70% of young internet users in China experienced emotional dependence on AI, and 23% rely on it regularly or habitually. In a survey conducted last year by the China Youth and Children Research Center with 8,500 minors, more than 20% responded that "they only want to talk with AI and not with real people."

In response, authorities implemented the "AI Partner Service" regulation on July 15, 2026. The law prohibits encouraging users' emotional dependence and bans virtual romantic relationships with minors, also requiring mandatory government safety checks before launch. Providers must clearly indicate content generated by AI, notify users every two hours that they are interacting with an AI, and inform emergency contacts if severe emotional crises are detected. Authorities reserve the right to suspend services deemed dangerous at any time.

Following these regulations, companies such as ByteDance's Doubao, Alibaba's Qiyuan, and MiniMax's Xingyan discontinued their customized virtual friend and virtual lover generating services. As a result, users who suddenly lost their AI avatars of virtual romantic partners, friends, or deceased family members expressed grief and depression on social networking sites (SNS). According to foreign media outlet "Rest of World," SNS has seen a flood of memorial posts and movements calling for the return of such services.

The New York Times analyzed that these regulations represent an attempt to strike a balance between the twin goals of spreading AI and maintaining social stability, noting it as "an effort to balance the ambition to promote widespread use of AI with the desire to avoid anything that could disrupt social order."

This content was produced with the assistance of AI translation services.