Emergency managers quietly use AI chatbots to draft disaster alerts, experts say

The article describes an Alabama emergency manager who used a chatbot to create a graphic about a "sting jet" wind threat from Hurricane Isaias; the graphic was meteorologically inaccurate. Experts including Laura Myers of the University of Alabama and Jeannette Sutton of the University at Albany say unofficial use of tools such as ChatGPT and Gemini is already widespread. Startup EM1, which has contracts with about 100 government agencies and school districts, has trained a model on Sutton's alert templates that it says matches or exceeds human performance on benchmarks, though none of its clients yet use it to automate live extreme-weather alerts.

Skepticism remains. A public-employee union in Syracuse, New York, sued over the city's use of AI for non-emergency 911 calls, and consultants caution that AI-generated messages may feel inauthentic to residents. Researchers also warn that easier alert production could worsen "warning fatigue." For now, the consensus is that AI should stay in a supporting role with a human expert in the loop.