Emergency managers are increasingly exploring generative AI to draft public alerts, summarize conditions and coordinate disaster information, even though many agencies do not formally permit its use. The appeal is straightforward: small teams must translate technical updates into clear instructions quickly, and human-written alerts often contain jargon, missing context or distracting detail.
A recent incident shows the risk. According to an emergency-management researcher interviewed by WIRED and Grist, a local official used a chatbot to create a graphic about a possible “sting jet” threat after Hurricane Isaias reached the Gulf Coast. The image supported the official’s intended local warning but was not meteorologically accurate. Publishing a persuasive error during an emergency can damage the trust that agencies need people to follow evacuation or safety instructions.
AI may still help with first drafts, translations and message templates, but the source describes a need for trained humans to verify facts and approve the final alert. It also notes that agencies vary in staff and technical capacity. The useful standard is not whether a message sounds clear; it is whether the underlying hazard data, location, timing and requested action have been checked against authoritative sources before the public sees it.