Seattle Deployed Corti AI on 911 Calls Without Public Disclosure

Since December 2023, the Seattle Fire Department has been running AI from Copenhagen-based startup Corti on every incoming medical 911 call, prompting dispatchers in real time to route certain callers to a nurse consultation line in Texas rather than dispatch an ambulance. As The Next Web reported in June 2026, the department never disclosed the system to the public, never submitted it for review under Seattle's surveillance ordinance, and never sought city council approval.

The system is framed as dispatcher support rather than automated decision-making; SFD Assistant Chief Chris Lombard said dispatchers retain final authority. But the AI's influence appears measurable: correcting an earlier claim by SFD Medical Director Michael Sayre, a department spokesperson put the increase in nurse line routing at 32% since live AI prompting began, though neither figure has been independently verified.

Seattle's 2017 surveillance ordinance, codified as SMC 14.18, requires city departments to get council approval before deploying technology that 'observes, monitors, or collects data about individuals.' The fire department's position is that Corti doesn't qualify because it doesn't store audio or identify callers. Ryan Calo, a law professor at the University of Washington and co-director of the university's Tech Policy Lab, offered a different view: 'A person who is erroneously routed outside of the 911 environment has a right to know how it happened.'

The human stakes of nurse line diversion are not hypothetical. Pamela Hogan, 71, called 911 in April 2022 reporting knee pain, was routed to the nurse line, waited more than 10 hours for a callback, and was found dead in her home weeks later. Her death predates Corti's live AI prompting by more than a year, and the nurse line existed before the AI was involved, but the case illustrates the consequences when callers are diverted from emergency response. Emily M. Bender and Decca Muldowney, writing on Buttondown, raise a further concern: 'automatic processing of speech...does not work equally well for different people,' meaning callers with regional accents, second-language accents, or speech disabilities may face different outcomes from a system making real-time triage recommendations.

The reporting doesn't provide outcome data on whether AI-prompted diversions track against patient results, or what the demographic breakdown of routed calls looks like. The fire department's interpretation of SMC 14.18 is now a question for the city council or the courts. The AI has been operating for 18 months. The public found out this week.