A Chinese farmer destroyed 25 acres [1] of sesame seedlings [3] after following pesticide advice generated by an artificial intelligence tool [1].
The incident highlights the risks of relying on generative AI for specialized agricultural chemical mixtures, where a single hallucination can lead to total crop failure.
The farmer, who is 67 years old [2], had used the AI tool for several months to manage his land [1]. Reports said the farmer had previously received successful recommendations from the system, which built a level of trust in the technology [1].
This trust led the farmer to apply a specific pesticide recipe provided by the AI for weed and pest control [1]. The recommended mixture included high-efficiency flupyrimethalin, flusulfasulfaether (flufenacet), thiamethoxazine, and methyl salt [4].
While the AI had provided helpful guidance in the past, this specific combination proved harmful to the sesame seedlings [1]. The resulting chemical reaction killed the entire 25-acre [1] crop, leaving the farmer with a total loss of his seedlings [1].
Agricultural experts said AI tools often struggle with the precise chemical interactions required for pesticide application. Because these models predict the next likely word rather than calculating chemical safety, they can produce recipes that look professional but are toxic to specific plant species [1].
The farmer's experience serves as a cautionary example of "automation bias," where users stop verifying information once a system has proven reliable a few times [1]. In this case, the transition from successful general advice to a specific chemical formula resulted in the destruction of the farm's output [1].
“A Chinese farmer destroyed 25 acres of sesame seedlings after following pesticide advice generated by an artificial intelligence tool.”
This event underscores the danger of applying generative AI to high-stakes physical environments without human expert verification. While AI can optimize logistics or general scheduling, its tendency to 'hallucinate' plausible-sounding but factually incorrect data can lead to irreversible material damage in sectors like agriculture and medicine.
