New Orleans has become the first major city in the United States to put artificial intelligence on the front lines to field emergency calls.
The Orleans Parish Communication District, which handles more than a thousand emergency calls every day in one of the busiest systems in the country, is now routing some of those calls through an AI system. That system determines which calls need a human dispatcher right away and which ones the bot can handle on its own.
The tool is called AI Emergency Call Triage, and it went live in late July — after the district first tested it on the non emergency 311 line. According to local reporting, the AI evaluates each call and filters out the urgent ones so human dispatchers are not flooded with redundant reports or lower priority issues.
Karl Fasold, the executive director of the Orleans Parish Communication District, has been clear that the AI is not meant to replace people. He said the system is mainly used in two situations: when a caller is within about 200 meters of a crash that has already been reported, and when every human call taker is already busy.
One online commenter — who goes by the name Boxy and says he spent 11 years as a 911 dispatcher — called the move “the worst idea you could possibly think of.” He pointed to the kinds of calls that are hard even for experienced humans, such as those made by people who cannot speak clearly or callers with strong accents, or even someone who might be in the midst of a domestic situation and pretending to order pizza because they are afraid to say what is really happening.
In those moments, Boxy argued, a dispatcher has to hear the tone and notice what is not being said. “You need a person,” he wrote.
New Orleans appears to be the first major city to let AI actually answer the emergency line. Other places have used the technology in more limited ways. Atlanta, for example, has begun to use AI to help live dispatchers to pinpoint caller locations and access relevant medical data.
Seattle began using a similar system in 2024 to help decide which calls need a rapid response. That program ran for more than two years before the public learned about it, according to an investigation by the Seattle Times.
While AI can often be an invaluable tool when used properly, studies have shown that AI voice systems can invent information at a high rate. One report from February found that tools similar to ChatGPT and Gemini delivered false details nearly half the time while sounding convincing.
People who dial 911 are rarely calm or easy to understand, which only increases the possibility of mistakes being made. Security researchers have also shown that voice models can be manipulated from the outside. Teams from Zhejiang University reported success rates between 79 and 96 percent when they tried to change how the systems behaved.
Officials in New Orleans say they believe the system can work if it stays under human oversight, is trained carefully, and is protected with strong cybersecurity and blunt rules for the algorithms.
.png)
.png)
