1. News
  2. AI models & agents
  3. AI · Emergency alerts
AI · Emergency alerts2 min read

Emergency officials are quietly using AI chatbots to write disaster alerts

Disaster officials may already be using AI chatbots to draft public warnings, and one county's AI-made storm graphic was wrong. Here is what that means for you.

By the Dashline deskSource: Wired
Emergency officials are quietly using AI chatbots to write disaster alerts
Photo: Wired

Local emergency officials are likely already using AI chatbots to help write disaster alerts, even though most agencies do not allow it. Experts say the tools could make warnings clearer. But a recent hurricane in Alabama shows the risk: an AI-made graphic that looked helpful was not accurate.

What happened in Alabama

After Hurricane Isaias hit the Gulf Coast as a Category 3 storm, a county emergency manager in Alabama published a graphic showing a possible "sting jet" of high winds. He had made it with an AI chatbot. Laura Myers, who directs the University of Alabama's Center for Advanced Public Safety, heard about it from a meteorologist. She said the tool did what he asked, but it was not meteorologically accurate.

Why alerts go wrong

Experts say the problem is not the officials. It is that they are human. They often use jargon, skip details they assume people know, and add facts that distract from what matters. Many agencies are short on staff and time.

Bad wording has real costs. In 2016, neighboring California counties issued a voluntary and a mandatory evacuation for the same wildfire. In 2017, a confusing notice before a mudslide in Montecito, California, told people in a voluntary zone to be ready to leave. The mudslide killed 23 people and injured more than 160.

Why AI is tempting

A chatbot can read huge numbers of past alerts and research on what works, then edit a draft in seconds. Rob Dale runs emergency management in Ingham County, Michigan. He said that in a crisis, with phones ringing, AI would probably do a better job than a person who last sent an alert 15 years ago.

Dale has used AI only for practice alerts. He said he would let a chatbot draft a real one, but he would still control sending it. A startup called EM1 says its AI model, trained on research by disaster scholar Jeannette Sutton, is already better than almost any human emergency manager. The company has tested it on scenarios such as wildfires and tsunamis.

What to do

  • Check any major alert against your local weather service or county website.
  • Look for clear actions, such as who should leave, where to go and by when.
  • Be cautious with storm maps or graphics that do not come from an official forecaster.

What is not known yet

No one has said how widely AI is used in alerts today. Experts believe it is common, but most agencies ban it, so use is likely out of view. The material also does not say whether the EM1 claims have been checked by outside experts.

The takeaway

Emergency alerts may soon be drafted with AI help, so confirm anything important with your local weather service or county site.

More articles

You should really read these.

Next step

Let’s talk