AI Chatbot's Pesticide Advice Destroys Chinese Farmer's 25-Acre Sesame Crop

Chinese farmer Wu, 67, had grown accustomed to relying on an unnamed AI application for agricultural guidance—asking about fertilizers and crop protection after early suggestions seemed to work. That confidence led him to follow the chatbot's recommendation for treating weeds and pests across roughly 150 mu (24.7 acres) of sesame. The AI suggested mixing multiple chemicals, including flupyrimethalin, flusulfasulfaether, thiamethoxazine, methyl salt, haloxyfop-P-methyl, and fomesafen. Wu applied the mixture without cross-checking with experts. By the next day, the treatment had killed not just the weeds but all the sesame seedlings, turning a crop-protection measure into a total field failure.

When Wu returned to the AI for an explanation, the system did what many chatbots do—it confidently blamed flusulfasulfaether. Yet those chemicals are legitimate plant-protection agents, particularly for soybeans; the issue was likely their combined use. This incident mirrors other AI trust failures, such as a 60-year-old man who developed a rare condition after ChatGPT told him to replace table salt with sodium bromide. The pattern is clear: AI can build trust through repeated correct answers, then hallucinate with equal confidence on a critical decision. Wu's case is a stark reminder that AI advice should be a starting point, not a final authority—especially when real-world consequences are irreversible.