A misleading headline made people think AI was replacing 911 dispatchers

[Skye Jacobs]

Sounding off: As AI moves into more parts of daily life, the possibility of using it in New Orleans' 911 system struck a nerve. Residents were relieved to learn that reports of AI replacing emergency call-takers were inaccurate, but the reaction underscored a larger reality.The technology is reshaping essential services, often faster than the public is ready to accept.

New Orleans is using artificial intelligence in a limited part of its 911 operation: screening some duplicate calls about vehicle crashes when human call-takers are unavailable.

The Orleans Parish Communication District issued the clarification after a Shreveport Times story reported that the city would use AI to answer 911 calls instead of human operators. OPCD said the technology is not replacing human call-takers and is used only in a narrow set of circumstances.

The Shreveport Times article, published Thursday under the headline "New Orleans will use AI to answer 911 calls instead of a human," focused on Carbyne's AI Emergency Call Triage tool. The system is designed to identify when multiple callers are reporting the same incident.

Under that process, callers reporting a crash that is already known to dispatchers may receive information or updates about the incident. Callers reporting a separate emergency would be routed to a human dispatcher. The story also noted that the tool would not replace dispatchers.

Still, the headline drew a wider reaction online, with commenters questioning whether AI should play any role in emergency communications.

"I wonder what happens if someone has to be discreet and order a 'pizza' and instead AI answers," one commenter wrote.

Some commenters also pointed to McDonald's abandoned test of AI drive-thru ordering. Widely shared examples of incorrect orders became a symbol of the technology's shortcomings, and McDonald's ended the test at more than 100 restaurants in 2024 after working with IBM on the automated ordering system.

The comparison isn't exact. Ordering food at a drive-thru is very different from calling 911, and OPCD says its system is limited to duplicate reports of a known crash. But the reaction reflects a broader concern about using AI in situations where mistakes could have serious consequences. Questions about reliability, accountability, and access to a human operator are particularly sensitive in emergency response.

OPCD Executive Director Karl Fasold said the agency has used the AI tool for about two years. It can be activated only when all human call-takers are busy and a caller is within about 200 meters of a crash that has already been reported.

The system then determines whether the caller is reporting the same crash. If so, it can provide information or updates. If the caller is reporting something else, the call is routed to a human dispatcher.

OPCD said the technology is not used for other types of 911 calls. The agency also said it has safeguards in place to ensure callers can reach a person when needed.

Fasold said the Shreveport Times article left readers with the impression that New Orleans was moving toward an AI-run 911 system. He said that is not how the city is using the technology.

OPCD also said the reporter did not contact the agency before publishing the article and did not identify sources for some of its claims.

"Every 911 center in the country is dealing with the reality of tighter budgets and inability to reach full staffing even with funding to do so. We have found this AI triage tool, used specifically for the scenario outlined above involving motor vehicle accidents, to be an extremely positive use of the technology to make the best use of our staff's very limited time," Fasold said in a statement.

Permalink to story:

A misleading headline made people think AI was replacing 911 dispatchers

 

[Cyclone7d]

So in the future when we call 911 in the U.S. we could get the line from 2001 A Space Odyssey, "I'm sorry Dave but I can't do that." I'm looking forward to this and it's probable upcoming tie in with all the Flock type spy cameras being used in the U.S. No hard facts about my last sentence, I can just use my personal Magic 8 Ball to see where this could go!

 

[Megalomaniac]

Press 1 for murder
Press 2 for rape
Press 3 if this is a White Collar crime including tax evasion, embezzlement or credit fraud
Press 4 if this is a suicide attempt
Press 5 if this is related to Trump’s economic policies

 

[brucek]

I'm pretty surprised McDonald's couldn't make order taking work for a limited menu (or maybe it was more that they just couldn't make it be more affordable than a minimum wage worker).

The transcription provider I work with in a much wider field has gotten pretty reliable including multiple languages, distinguishing different speakers, etc. Although I don't need real time conversation so maybe that is part of what made it harder.

 

[JMGuy]

It is disheartening that the primary output of our free press has become weaponized panic designed to farm engagement and control narratives. Too many people have stopped doing the basic mental lift of critical thinking, choosing instead to buy into whatever one-dimensional outrage headline crosses their feed and regurgitating it as if that’s some kind of signal toward intellect.

True substance requires stepping outside of a comfortable echo chamber. If you only surround yourself with people who agree with you, you become incapable of any actual intellectual or emotional growth. "Being right" has very little value compared to the actual discovery of learning where you’re wrong.

When sensationalism pays better than nuance, and vetting the complete truth gets treated like a secondary luxury, it becomes crystal clear why public discourse feels so completely broken.

 

[Squid Surprise]

I love how we compare AI to a human who is always 100% accurate… AI is improving rapidly… humans are not… while most humans can do select jobs better than some AIs now -that won’t be the case for much longer.

People need to stop equating AI with Skynet and realize that it’s not evil, won’t nuke Russia, and will make our lives far better in the very near future - if it isn’t doing so already.

 

[yRaz]

I love how we compare AI to a human who is always 100% accurate… AI is improving rapidly… humans are not… while most humans can do select jobs better than some AIs now -that won’t be the case for much longer.

People need to stop equating AI with Skynet and realize that it’s not evil, won’t nuke Russia, and will make our lives far better in the very near future - if it isn’t doing so already.

While I can't remember it off the top of my head, there is a term for how good an AI can be relative to how much data it has and processing power it has available to it. We can simulate intelligence with inferencing and next is to give it the ability to reason. Creating and powering just one of these machines that can accurately simulate a human would require so much compute it might actually be impossible to build without quantum computers. And that's just for one "agent". What stuck with me is the phrase "there isn't enough sand on earth to replace all humans with AI". This was midn2025 and things are different now, but progress has definitely slowed and building bigger datacenters hasn't turned into the massive leaps in AI that have been talked about. Where does this end, terawatt data enters will billions of GPUs. To replace a single human? I can reliably get 5-6 props out of an AI and it's 50/50 if I can get passed 10. Sometimes that's all you need, but not to replace someone on an 8 hour shift with thousands of interactions.

 

[Squid Surprise]

While I can't remember it off the top of my head, there is a term for how good an AI can be relative to how much data it has and processing power it has available to it. We can simulate intelligence with inferencing and next is to give it the ability to reason. Creating and powering just one of these machines that can accurately simulate a human would require so much compute it might actually be impossible to build without quantum computers. And that's just for one "agent". What stuck with me is the phrase "there isn't enough sand on earth to replace all humans with AI". This was midn2025 and things are different now, but progress has definitely slowed and building bigger datacenters hasn't turned into the massive leaps in AI that have been talked about. Where does this end, terawatt data enters will billions of GPUs. To replace a single human? I can reliably get 5-6 props out of an AI and it's 50/50 if I can get passed 10. Sometimes that's all you need, but not to replace someone on an 8 hour shift with thousands of interactions.

Any “rule” we have about AI is constantly changing as our tech does.
And we don’t need to perfectly replace a human - we need to replace the work they do over a specific period of time. If that work is something like working a call center - it can already do that pretty well. Remember that the average human isn’t that bright - and half are dumber still

Most call-centre employees work off of a script anyways - an AI can be given a far more complex one and react instantly unlike a human who will put you on hold while they leaf through their manual…

Give it a few more years and most people will WANT the AI when they call in…

 

[hahahanoobs]

Just goes to show the average consumer knows nothing about AI or common sense.