New Orleans Deploys AI to Answer 911 Calls, Raising Questions About Emergency Response Safety

Better Hope You Don't Need an Ambulance in New Orleans, Because the City's 911 Operators Are Now AI

Can't-miss innovations from the bleeding edge of science and tech If you thought AI customer service bots were bad, just wait until you need to convince a large language model that yes, you have been involved in a car accident, actually. According to new reporting by the Shreveport Times, officials at the Orleans Parish Communication District -- the emergency dispatch agency for the city of New Orleans -- are now integrating AI answering agents into its workflow. The new tool is called AI Emergency Call Triage, offered by a company called Carbyne. Its function is essentially to triage incoming calls to dispatchers, bumping high-priority emergency calls to humans, and automatically handling repeat calls related to ongoing incidents, the Times explains. Starting in April, New Orleans experimented with implementing the AI for 311 calls, a phone line meant to connect callers to local government information and non-emergency services. That the AI will now be sorting incoming emergency calls is a significant escalation for the city's dispatch agency. New Orleans will now join other major cities including Seattle and Atlanta, which have also implemented AI into their emergency dispatch services in recent weeks. Though officials claim the new tech will help cut through the hundreds of calls inundating human dispatchers, it nonetheless represents an added layer of friction between callers in distress and trained dispatchers who can connect them with help. It's also worth noting that AI voice chat remains highly flawed even when powered by frontier AI models, with plenty of opportunities for errors both in how they process information, and how they dish it out. A February study by the media software company NewsGuard found that in voice mode, AI bots like ChatGPT and Gemini produced false claims with realistic audio up to 50 percent of the time. Emergency calls to 911, needless to say, are often extremely chaotic. Meanwhile, AI voice systems are also persistently vulnerable to outside attacks, with cybersecurity researchers successfully manipulating voice model's behavior with an average success rate between 79 and 96 percent. What all of that means for situations where connecting with a human dispatcher could mean the difference between life and death remains to be seen -- but if other AI voice experiments are any indication, it won't be good. More on first responders: Emergency Responders Say They're Now Unpaid "Roadside Assistance" for Confused Waymos

AI Is Already Answering Some of Your 911 Calls

AI has worked its way into daily life since ChatGPT made it a household name in 2022. It writes emails, edits photos, and recommends what to watch next. Now add another place it's been quietly operating: the 911 call center in New Orleans. Orleans Parish officials confirmed this week that artificial intelligence has been handling a slice of the city's 911 calls. What the AI actually does The AI only handles a narrow type of call. It kicks in when someone dials 911 near a car accident that's already been logged in the city's dispatch system, specifically within 200 meters. If the AI confirms the caller is reporting that same accident, it tells them help is already on the way. It then asks them to stay on the line only if they were involved or have new details. Everyone else can hang up. Orleans Parish Communications District director Karl Fasold says the tool exists for a simple reason. Rush-hour pileups used to flood the phone lines with dozens of calls about the exact same crash. That pushed more urgent emergencies further back in line. According to the district, the AI now handles 15 to 18 out of every 20 of those repeat crash calls. Officials are clear that violent crimes and life-threatening emergencies never go to the AI. For those particular calls, a human will always be the first voice you hear. Why some experts want more guardrails Not everyone is fully comfortable with the setup. Tulane University professors who study AI in public services point out something important. There's currently no city or national policy that specifically governs how AI gets used in emergency response. Callers also aren't explicitly told they're speaking to a machine. One of those professors, Rob Lalka, states it quite plainly. Get it wrong even once, he said, and someone's life could be on the line. New Orleans is also getting ready to bring a similar AI tool to its non-emergency 311 line.