Every 911 medical call placed in Seattle since December 2023 has been run through an AI system that dispatchers never publicly disclosed, according to a 2026 report examining the Seattle Fire Department’s use of technology from the Danish company Corti. The AI listens in on calls and prompts dispatchers in real time — sometimes suggesting that a caller be routed to a nurse advice line instead of receiving an ambulance.
How the Relationship With Corti Began
Seattle Fire’s relationship with Corti dates back further than most residents would guess. The department and Corti signed an initial data-sharing agreement in 2019, and Corti began testing cardiac-arrest detection algorithms on Seattle’s call data even before the pandemic. What started as a research partnership escalated in December 2023, when the department began actually acting on Corti’s live prompts during real calls — a shift from passive research tool to an operational system shaping which patients get an ambulance and which get redirected elsewhere, according to SFD Assistant Chief of Resource Management Chris Lombard and Medical Director Michael Sayre.
The Numbers at the Center of the Dispute
Corti has claimed its system increased the rate of calls diverted to a nurse line by roughly 50%; the fire department disputes that figure and puts the real increase closer to 32%. The department pays roughly $260,000 annually for the technology. Those numbers matter because a diversion means someone who called 911 believing they needed an ambulance instead gets a phone consultation — a decision with real consequences if the triage is wrong.
A Death That Raised the Stakes
The report’s most consequential detail involves Pamela Hogan, a Seattle retiree who was diverted to a nurse line during a 2022 call — before the AI prompts were formally adopted for live use, but during the period Corti was already influencing department practice. Hogan waited more than ten hours for an ambulance and was later found dead. Her estate is now suing the city, and the case has become the focal point for arguing that AI-influenced triage decisions carry life-or-death stakes that deserve public scrutiny before deployment, not after a lawsuit forces disclosure.
Why Nobody Was Told
Seattle has a 2017 surveillance ordinance requiring city council review and approval before departments deploy technology that observes or analyzes members of the public. The fire department determined on its own that Corti’s system didn’t meet that threshold, reasoning that the AI doesn’t store audio recordings or identify individual callers by name. University of Washington law professor Ryan Calo has publicly challenged that self-assessment, saying “the potential [for harm] raises serious concerns” and arguing the AI appears to be actively prompting dispatchers to push calls toward the nurse line rather than passively assisting them. Lombard has defended the department’s practice by emphasizing human oversight: “the dispatcher still has ultimate authority” over any diversion decision, he said, meaning the AI is designed as a recommendation layer rather than an autonomous decision-maker.
The Case for the Technology, Made by Its Backers
Supporters of AI-assisted dispatch argue the underlying problem is real and worth solving: 911 systems nationwide are strained by call volume, and predictive tools that flag likely cardiac arrests or route lower-acuity calls to nurse lines can free paramedics for the emergencies that most need them. Industry analyses of AI in EMS point to automated transcription, translation, and administrative support as ways to let dispatchers focus attention on listening and assessing rather than paperwork, and predictive analytics that anticipate call-volume patterns to position ambulances more efficiently. The argument isn’t that AI triage is inherently unsafe — it’s that going without public disclosure removed the chance for the community, and the city council members like Bob Kettle who oversee surveillance policy, to weigh those tradeoffs before the system went live on real emergencies.
What’s Next
Seattle officials have said they are now developing a public-facing governance framework to retroactively assess AI tools like Corti’s against privacy, accountability, and public-interest standards — an acknowledgment that the current review process failed to catch a system that had been running on every medical call for more than two years. The Hogan lawsuit will likely test in court whether “the dispatcher retained ultimate authority” is a sufficient legal shield when an algorithm’s recommendation preceded a fatal delay. For other cities watching Seattle’s example, the case is becoming a cautionary tale about deploying AI in life-safety systems quietly, informing the next round of state and municipal rules on how emergency AI must be disclosed and audited before it goes live rather than after something goes wrong.
