New Orleans has become the first city in the US to use artificial intelligence to answer emergency 911 calls — but critics warn it could have deadly consequences.
The Orleans Parish Communication District — which gets over a thousand emergency dials a day, in what is among the highest volume of calls in the country — is routing the 911 calls to an AI agent that decides which ones get bumped to human dispatchers and which get handled by the bot itself, according the Shreveport Times.
“I was a 911 dispatcher for 11 years and this is the worst idea you could possibly think of,” warned an online critic using the name Boxy, who claimed to be a first responder.
The new tool, known as AI Emergency Call Triage, was first deployed in Louisiana’s largest city in late July after being tested on its 311 hotline.
The triage bot analyzes calls to prioritize which emergencies should be handled by humans, while it handles repeat calls or low-priority ones to avoid inundating dispatchers, the Times explained.
OPCD Executive Director Karl Fasold, however, emphasized that the AI is not replacing human dispatchers, and said the system will mainly be used when the caller is within roughly 200 meters of a crash that has already been reported and when all human call-takers are busy, according to local CBS News affiliate WWL-TV.
Critics, nonetheless, fear it could have deadly consequences.
“You have scenarios where the person can’t speak well [or have strong accents] — or if they’re in a domestic situation and they pretend to call for pizza or something — you have to be able to determine how someone feels or pick up on what they aren’t saying,” said Boxy, who had a picture of himself in a first responder uniform in his profile.
“You need a person.”
New Orleans appears to be the first major municipality in the country to allow AI to actually answer emergency calls.
However, others — such as Atlanta and Seattle — have also integrated AI into their 911 systems in lesser forms.
Atlanta uses AI to help live dispatchers locate callers faster, according to CBS News.
Meanwhile, Seattle started using a similar tech in 2024 to help dispatchers decide which callers deserve a rapid response and which don’t — using it for over two years without disclosing it to the public, according to an investigation by the Seattle Times.
But AI’s track record on accuracy, especially when it comes to voice chatting, is sketchy at best, according to multiple studies.
In February, one study found that bots like ChatGPT and Gemini made up false information nearly half the time, delivering it in audio that sounded convincing, according to data collected by Newsguard.
Complicating it more is the fact that people calling 911 are rarely calm, coherent or easy to understand. Critics warn that only increases the risk for misunderstanding.
AI voice systems are also persistently vulnerable to outside attacks and could even develop hidden biases based on the data it’s fed, with cybersecurity researchers successfully manipulating the voice model’s behavior with an average success rate between 79% and 96%, according to security researchers from China’s Zhejiang University, presented during May’s IEEE Symposium on Security and Privacy.
Security researchers have also managed to hack these systems from the outside, tricking voice models into misbehaving with a success rate as high as 96% in some tests.
Still, officials in New Orleans said they strongly believe that with human oversight, thorough training data, strong cybersecurity and algorithm regulation, the AI 911 bot could work effectively.
Read the full article here
