Seattle used AI to monitor 911 medical calls for years in secret

The Seattle Fire Department has been using artificial intelligence to monitor every 911 medical call placed in the city since December 2023, all without any public disclosure or formal review under the city’s surveillance ordinance. According to an investigation by GeekWire, the AI system, developed by Copenhagen-based startup Corti, analyzes calls in real time and prompts dispatchers to redirect certain callers away from emergency responders and toward a nurse consultation line operated out of Texas. The department never informed the public, sought city council approval, or submitted the technology for evaluation under Seattle’s surveillance oversight law.

Seattle’s surveillance ordinance, codified as SMC 14.18, mandates that any city department must obtain council approval before deploying technology that “observes, monitors, or collects data about individuals” in ways that could raise civil liberties or social justice concerns. Passed in 2017, the ordinance was designed to give residents a voice in how the city uses surveillance tools. A Seattle Fire Department spokesperson told GeekWire that the department does not classify Corti as a surveillance technology because it neither stores call audio nor identifies individual callers.

Ryan Calo, a law professor at the University of Washington and co-director of the university’s Tech Policy Lab, challenged that interpretation. “A person who is erroneously routed outside of the 911 environment has a right to know how it happened,” Calo told GeekWire. The ordinance does not explicitly distinguish between surveillance and decision-support, and the question of whether an AI system that listens to emergency calls in real time qualifies as surveillance technology remains legally untested.

Corti raised $60 million in a Series B funding round and works with emergency services across multiple countries. The Seattle Fire Department initially partnered with Corti in 2021 for triage support, but the live AI prompting feature that routes callers to a nurse line began in December 2023.

SFD Medical Director Michael Sayre claimed in a 2024 Corti press release that nurse line routing had increased by 50% since the AI was deployed. An SFD spokesperson later corrected that figure to 32%, though neither number has been independently verified. The discrepancy between the department’s own figures highlights a lack of external oversight.

The nurse line itself is not a new concept. Seattle has used nurse consultation lines for years to manage lower-acuity 911 calls, freeing up ambulances and paramedics for emergencies. What changed in December 2023 is that an AI system began actively prompting dispatchers to divert calls there.

SFD Assistant Chief Chris Lombard emphasized that dispatchers retain final authority over routing decisions. However, the presence of AI prompting in the 911 call workflow introduces an automated layer of judgment that callers are unaware of.

The case of Pamela Hogan illustrates the stakes. Hogan, 71, called 911 on April 8, 2022, reporting knee pain. She was routed to the nurse line, waited over 10 hours for a callback, and was found dead in her home weeks later. Her estate is suing the city, and the lawsuit cleared a dismissal hurdle earlier this year.

Hogan’s death occurred more than a year before Corti’s live AI prompting began, and the nurse line that failed her existed before the AI was involved. But the case demonstrates the consequences when callers are diverted away from emergency response. It raises the question of whether an AI system that increases the rate of such diversions should face public scrutiny before deployment.

The lack of disclosure is part of a broader pattern. Seattle Mayor Katie Wilson said her administration is developing a public-facing AI governance framework that will “center human flourishing and serving the public good.” Wilson previously paused the expansion of Microsoft Copilot tools within city government, signaling awareness that AI deployments require oversight.

But the Corti system was already running for months before any governance framework was announced. There is no indication that the fire department sought or received mayoral approval for the deployment.

The technology is spreading beyond Seattle. Snohomish County uses a different AI system called Ava, built by a company called Aurelian, for non-emergency calls, and says it has handled more than 220,000 of them. Kitsap County has also deployed Aurelian’s system on a new non-emergency line.

These deployments differ from Seattle’s in an important respect: they are limited to non-emergency calls. Corti listens to all 911 medical calls, including those that may involve life-threatening emergencies.

The growing push for AI oversight in the United States has focused primarily on frontier models and national security applications. Municipal AI deployments in emergency services have received far less attention, even though they affect people in moments of acute vulnerability. The question Calo raised, whether someone routed away from emergency response deserves to know an AI was involved in that decision, does not have a clear legal answer in most jurisdictions.

Corti’s system is not making the final call. Dispatchers can override the AI’s prompts, and SFD says they do. But the 32% increase in nurse line routing suggests the AI’s recommendations are being followed more often than not.

When public institutions deploy AI in healthcare settings, the transparency gap between what the institution knows and what the public knows becomes a governance problem, not just a technical one. Seattle built a surveillance ordinance to close exactly that kind of gap. The fire department decided the ordinance did not apply.

Whether that interpretation holds up is now a question for the city council, the courts, or both. The AI has been listening for 18 months. The public found out this week.