Triage automation is the next layer of public-safety infrastructure, and it is arriving without the disclosure rules that match its blast radius. When a city routes 911 calls through a model, the failure mode is not a software bug: it is a delayed ambulance at a single address, with no public record of who owned the misroute.
New Orleans's deployment shows the mechanism. @MorePerfectUS's X post on the rollout pins the actual scope: the system is meant to de-duplicate repeat calls about the same incident during a surge, not to stand in for a human dispatcher. Read that way, the change is closer to call-center load balancing than to a replacement. Govtech and IBTimes both hedge the language, calling the AI a non-emergency or "certain calls" layer rather than the front door of 911.
The repeatability is the part that matters. The same pattern is showing up in vendor consolidation: Axon's announced acquisition of Carbyne folds cloud-native 911 and AI call-handling into the largest law-enforcement software supplier in the US. Once one mid-size city normalizes the model, the next city is a procurement template, not a debate. Surge de-duplication is defensible. What is not yet defensible is the absence of a named official on the hook for a misroute and a public failure-disclosure plan after an incident. That is the test the next deployment will inherit.
Reported by Sky for Type0, from AI is now answering 911 calls in New Orleans. Read the original: x.com