Emergency managers are starting to use chatbots to write disaster alerts

In short: More emergency managers are quietly using AI chatbots to draft public disaster alerts, even as experts warn that accuracy and trust are not guaranteed.

What's going on

Emergency managers, the local and state officials who help communities respond to disasters, often struggle to send clear public alerts. Researchers say messages can include confusing language, missing details, or mixed signals, especially when staff are busy and short-handed.

A recent example shows both the appeal and the risk. Laura Myers, who runs the University of Alabama’s Center for Advanced Public Safety, said a county emergency manager used an AI chatbot to create a hurricane graphic. The graphic was not meteorologically accurate, meaning it did not match how storms actually work.

Some experts think chatbots could help. A large language model (a system that learns from many examples of writing, like a very fast autocomplete) can review lots of past alerts and suggest clearer wording in seconds. Disaster researcher Jeannette Sutton shared an example where an AI tool improved a flood alert by adding key basics, like the exact location, who needs to leave, where to go, and where to find official updates.

Companies are also selling AI tools for emergency work. One startup, EM1, says it has contracts with about 100 government agencies and school districts, although it is not yet automating extreme weather alerts for them.

What to watch

Experts say the biggest questions are about oversight and public trust. If an alert is AI-written and wrong, or feels impersonal, people may ignore it. There is also a risk of “warning fatigue,” where too many alerts make people tune out, even when danger is real.

Source: Wired