AI Tools Quietly Draft Disaster Alerts for 100 Agencies as Accuracy and Trust Risks Persist
Updated
Updated · WIRED · Oct 28
AI Tools Quietly Draft Disaster Alerts for 100 Agencies as Accuracy and Trust Risks Persist
1 articles · Updated · WIRED · Oct 28
Summary
Emergency managers are already using ChatGPT- and Gemini-style tools to draft disaster messages unofficially, even though most agencies do not formally allow generative AI for public alerts.
Resource-strapped local departments see AI as a way to turn jargon-heavy, error-prone warnings into clearer instructions faster, a potentially life-saving advantage when officials rarely send alerts and struggle under pressure.
EM1 says its model—tested on wildfire and tsunami scenarios and sold to about 100 government agencies and school districts—can match or beat most human managers, though clients currently use it only for planning and internal preparation.
Accuracy remains a live risk: one Alabama emergency manager used a chatbot to create an inaccurate Hurricane Isaias “sting jet” graphic, showing how AI can confidently produce meteorologically wrong information.
Trust is the bigger unresolved question, with experts warning AI-generated alerts could feel less authentic, increase warning fatigue, and erode confidence in the public messages that can determine whether people evacuate in time.