Although emergency-management organizations are beginning to use artificial intelligence informally, government agencies and school districts have not yet adopted chatbots to automate extreme-weather warnings. Experts say AI could make alerts faster and more effective, but it could also undermine public trust, contribute to alarm fatigue and produce messages that inexperienced officials cannot properly evaluate.
EM1 has contracts with about 100 government agencies and school districts, but none is currently using the company’s technology to automate extreme-weather warnings. For now, the organizations use it for workshop planning and internal preparation. Still, every expert interviewed for this article said many emergency managers are already using chatbots to draft messages, even if they are reluctant to acknowledge it. Sutton said emergency managers who attend her alert-training sessions often show her high-quality alerts that they created in seconds with ChatGPT or Gemini.
“Official adoption rates are pretty low, but people use it informally, definitely,” said Sarah Miller, a professor of emergency management at Pierce College in Florida and a longtime instructor at the National Disaster Preparedness Training Center. “They’re either using their personal account or they’re using a free account.”
The potential tradeoff may seem straightforward: If AI can produce better alerts more quickly than humans, and the difference between an effective warning and a poor one can be the difference between life and death, governments might use it for every heat wave and wildfire. But even researchers who are optimistic about chatbot technology say the change could backfire. People who are skeptical of AI or already distrustful of government may reject institutions that publish AI-generated graphics or social-media posts. In Syracuse, New York, for example, a public-employee union sued the local government over its use of AI services to respond to non-emergency 911 calls.
“AI-generated content doesn’t feel authentic,” said Ginny Katz, founder of HasAdapt, a consultancy that helps disaster managers use technology. Chatbots could also enable emergency managers to issue more alerts in less time, potentially increasing the risk of alarm fatigue.
Human oversight remains essential
As in other fields where AI is being deployed, disaster experts say the software is currently only as effective as the people using it. Experienced emergency managers may be able to accelerate public communication by having a chatbot draft a series of alerts. Inexperienced managers, however, may not know how to evaluate output from models such as ChatGPT or Claude when they are rushing to produce emergency warnings.
Although the technology is far from being capable of operating without human experts, that has not stopped many understaffed emergency-management departments from using it. Experts say it is still too early to determine how the spread of AI-generated content will affect public trust in government.
“Which do you think is more accurate: a human sending a text message or a computer sending a text message?” said Justin Snare, founder of Preppr, another company that provides AI software to emergency managers.
His answer was that it depends on the person receiving the message.
Source: www.wired.com
