The people inside 911 spent decades building a foundation. Cell phones broke the original system, so they adapted. The internet created new problems, so they wrote standards. Shims and wedges went everywhere to make it all work. The goal was always the same: a network of networks, interoperable, reliable, built to last.
They were building the map. They just didn't see who was redrawing it.
It wasn't the people writing the standards who saw AI coming. It was people on the outside looking in. Solutions appeared from outside the system — voice translation, non-emergency call triage, video-based weapons detection. Each one solving a problem the legacy network couldn't. Each one arriving without asking permission, and without building on the foundation that had been so carefully laid.
911 became a data network not by decision, but by accumulation. Text to 911. OnStar sending crash data. Alarm systems. LifeAlert. The industry had always said voice is best. Probably it is. But technology changed the question before anyone had a plan for what came next.
The insiders built the foundation. The outsiders bypassed it. Now AI is running on top of it, and nobody owns what happens next.
AI is a clever solution. Trainable on data, and 911 has a lot of data. New entrants happily gobble it up, training models to do all sorts of things.
Training 911 professionals on scenarios and procedures. Prompting them in the midst of a crisis. CPR instructions when a child can't breathe and the dispatcher just realized it's their daughter's best friend.
Triaging calls from lonely old men who just need someone to talk to. Not because they are a nuisance, but because 911 centers get busy.
These are real problems. The technology is solving real things. That is not in question.
But there is a question that isn't asked often enough: how do we know what the AI is actually doing?
We hear it constantly. AI can't replace the dispatcher. It can't replace the humanity. It makes mistakes, it misses things. Anecdotally those are easy arguments. But how do we prove them? Or maybe the better question is the inverse: how do we prove the technology isn't making those mistakes?
Who is responsible when technology fails? The dispatcher? The 911 director? The vendor? How do we even audit it? Recreate the incident: 12 systems touched the call, 12 logs, nothing matches.
AI decided it wasn't an emergency. Just an old man needing a friend. His neighbor found him cold an hour later.
How did the AI decide? Who can explain it? Who is responsible?
Right now, the honest answer is: nobody knows. Because the infrastructure to answer that question doesn't exist.
What can actually be done? What does minding the foundation even mean?
AI was the unknown unknown. There was no plan for managing its growth into 911. That's not an accusation. It's the reality of how innovation works. You can't plan for what you can't see.
But we can go back to basics. Who knew what, when, and how. These are the same questions 911 has been asking for decades. The difference now is that we need consistent answers across systems, across networks, across technologies.
When 911 was a voice network, it was simple. Listen to the recording. Everything was laid out in sequence.
Now it's voice and video and text and an alarm and a camera feed and a drone, all logging a timeline to a different system. We need one record. One consistent answer to what actually happened.
Without it, the network of networks is just a bunch of jumbled wires and information. A concept that never became a foundation.
Closing it is the work. Not more technology on top of an untrusted base. The base itself.