A 67-year-old farmer in Chuzhou, China, turned to an AI chatbot for advice on controlling weeds and pests in his sesame fields. The chatbot recommended a combination of agricultural chemicals, and the farmer followed the instructions after the AI had previously provided useful advice. However, by the following morning, approximately 24.7 acres of sesame seedlings had died, turning what appeared to be helpful guidance into a costly mistake.
The problem centered on one of the recommended herbicides, which is primarily used to control broadleaf weeds in soybean fields. Sesame is also a broadleaf plant, meaning the chemical could damage or kill the crop itself. The farmer had additionally applied the product across the entire field, while the AI later indicated that it was intended for targeted application to affected areas.
The incident illustrates a major weakness of general-purpose AI advice: a chatbot can produce an answer that sounds confident and technically detailed while missing an important real-world context or safety consideration. The farmer initially distrusted the AI but became more confident after receiving advice that appeared to work. That successful experience may have encouraged him to trust the system with a much higher-stakes decision without independently verifying the recommendation.
The broader lesson is that AI should be treated as a decision-support tool rather than an unquestionable expert, particularly when mistakes can cause significant financial, health, or environmental damage. In agriculture, reliable AI systems should ideally use crop-specific information, local conditions, verified chemical labels, and expert validation before recommending treatments. The story highlights why AI adoption must be accompanied by verification, explainability, and appropriate human expertise—especially when the cost of a wrong answer can be an entire harvest.
