911 gets an AI assist: are bots a lifeline or a liability?

New Orleans is using AI to answer some 911 calls, but only for a narrow category of incoming calls within 200 meters of an already-reported vehicle crash and only when dispatchers are busy, according to the Orleans Parish Communication District. The district called a Shreveport Times article reporting that AI would replace human dispatchers 'highly inaccurate.' Northeastern University professor Stephen Wood said the AI augments rather than replaces dispatchers, and calls like 'I'm having chest pain' will always be routed to a human.

911 gets an AI assist: lifeline or liability? New Orleans is deploying AI to answer some of the city’s 911 calls, prompting questions about whether the technology will now be the new normal. “911, what is your emergency?” It’s a familiar question, but there’s a new twist: turns out that in some cases, a robot might be on the other line. In recent days, New Orleans officials made headlines for deploying AI to answer some of the city’s 911 calls, prompting questions about whether having bots handle 911 calls will now be the new normal. The news also sparked debates about what happens or doesn’t when a bot answers a call for help. It seems that all the buzz started with a provocative headline — about something that wasn’t exactly news in the first place. On July 29, 2026, the daily newspaper Shreveport Times reported that “New Orleans will use AI to answer 911 calls instead of a human.” However, the Orleans Parish Communication District issued a statement https://www.opcdla.gov/opcd-responds-to-article-on-use-of-ai/ saying that it “is deeply disappointed in the highly inaccurate article” posted by the news agency. The triage tool has been used for the last two years “only for a very narrow category of incoming 911 calls” originating “from within a 200-meter radius of an already-reported vehicle crash,” the statement read. It only comes into play when dispatchers are busy. And anyone who calls to report another incident — or one that falls into a category other than traffic — is routed to a human. While the AI dispatch system in New Orleans might be narrow in scope, it does seem that policing in the Big Easy is getting more high-tech. Jesse Vad, communications director of the American Civil Liberties Union of Louisiana https://www.laaclu.org/ , a nonprofit, told Northeastern Global News that while the use of AI for 911 isn’t currently on the organization’s radar, they’ve been keeping a close eye on a related trend: the use of police drones. If you think all of this might be asking for trouble, you’re not alone. Many have questioned whether outsourcing such a sensitive task to AI bots — no matter how well-trained — could be a risky call. Northeastern Global News talked to Northeastern University professor Stephen Wood https://bouve.northeastern.edu/directory/stephen-wood/ about the situation and his thoughts on putting bots on the receiving end of a 911 call. Here’s what he shared. Editor’s Picks Why put bots on the line in the first place — and what are they actually doing there? Right now, AI is only being used in a very specific scenario, Wood confirmed. “If a call comes in within 20 meters of a crash, it’s already been reported, and all the humans are busy — the AI takes it,” he said. The distinction is key: the bots aren’t the first on the scene, or even on the phone. They’re stepping in when a public emergency triggers a flood of calls about the same thing. Say you drive by a highway pileup. As a Good Samaritan, you might think: “I better call in case no one else has,” Wood explained. And while a human would handle the first of those calls, the next 30 might go to the bots. “It’s not replacing, but augmenting what they do,” Wood added, explaining that the system helps keep dispatchers from getting bogged down while more urgent calls — and lives — hang on the line. Will certain calls always get routed to a human? Absolutely — at least if everything goes as planned, Wood said. “I’m having chest pain” is a call that will always go to a human, he explained, adding that the hope is the system will spot the signs of a situation serious enough to warrant human attention. Besides, there’s more to the dispatcher’s job than simply picking up the phone, asking what happened and sending a vehicle out, Wood explained. Calls also don’t fit neatly into a box, and dispatchers have to get information, read between the lines, tease out missing details and sometimes piece together what’s happening from a caller who may not be able — or willing — to spell it out. Those are all tasks that belong squarely in the human realm. Could this system backfire? That’s the worry, Wood said, suggesting that the biggest risk comes with AI mistaking separate incidents for the same one — a mishap that’s hard enough to avoid without bots in the mix. Wood recalled an incident from his years as a MedFlight paramedic. A team was treating a critically injured person on one scene when a call came in from a nearby spot reporting a similar injury. Dispatchers first assumed it was the same incident. “But then someone put two and two together,” Wood explained. With helicopters already circling the scene, why would someone call 20 minutes later? Sure enough, it was a different crash. “It’s a nuance AI wouldn’t pick up on,” Wood said. Likewise, Mayu Tobin-Miyaji, a law fellow from the Electronic Privacy Information Center https://epic.org/ EPIC , a nonprofit public interest research center in Washington, D.C., told Northeastern Global News that while AI systems could help triage calls, limits are important. “Government agencies should ensure that it doesn’t lead to mission creep, such as deploying an AI system to handle 911 calls entirely or broadening surveillance systems, and limit data collection to what is necessary,” she said, suggesting that trouble starts brewing when AI tries to do more than what it’s capable of, or collecting more data than it needs. What about handling emotions on the other end? Missed distress cues are the biggest concern when it comes to outsourcing dispatching tasks to AI, Wood said, adding that it’s especially true in a domestic violence situation. It’s common knowledge that “if you call 911 to order a pizza, that’s a person reporting a domestic violence incident,” he said, adding that if the AI takes it literally and suggests the nearest joint, the results could be disastrous. The same is true for handling callers with suicidal ideation — another situation where certain words or phrases might carry meanings that AI could overlook. It could also miss what’s hiding behind the words, such as a fearful whisper or a caller struggling through tears. And it’s not just emotions, either — regional dialects can trip up the system, too, he said. For example, he explained that saying “This patient’s got the sugar” is a common way to signal a diabetes emergency. But would AI pick up on it? It’s doubtful, Wood said. All in all, “I just don’t think it’s there yet,” Wood said, referring to AI’s ability to handle more complex dispatching tasks.