Faris Adani
San Francisco’s emergency services are increasingly frustrated, finding themselves acting as de facto roadside assistance for Waymo’s self-driving vehicles, a situation city officials warn is rapidly escalating into a critical public safety concern that diverts vital resources and endangers citizens.
The escalating tension between San Francisco city officials and Waymo, the autonomous vehicle (AV) subsidiary of Alphabet, reached a boiling point during a recent public hearing where the city’s top emergency management official publicly decried the constant need for police and firefighters to intervene when Waymo robotaxis become confused or immobilized. Mary Ellen Carroll, executive director of San Francisco’s Department of Emergency Management (DEM), minced no words in her testimony, stating that the frequency of these incidents is no longer merely an inconvenience but a grave threat to public safety. Her concern stems from a pattern of Waymo vehicles obstructing traffic, including during critical incidents, thereby forcing highly trained first responders to spend valuable time and resources managing robotaxis instead of addressing genuine emergencies.
The issue was starkly highlighted by a city-wide power outage in late December, which plunged parts of San Francisco into darkness and, predictably, threw Waymo’s fleet into disarray. Traffic lights went dark, rendering the robotaxis’ navigation systems bewildered. Instead of safely pulling over or finding alternative routes, numerous Waymo vehicles simply froze in place, piling up at intersections and creating significant gridlock. Carroll recounted the chaotic scene, detailing how at least four major intersections became choked with stalled Waymos, requiring police officers to physically intervene. This intervention often involved a laborious process of contacting Waymo—a task Carroll claimed was often met with frustratingly long hold times, with one staffer allegedly waiting nearly an hour—or summoning tow trucks, and in some cases, officers resorted to manually moving the heavy vehicles themselves. This, she emphasized, is an untenable situation, transforming essential emergency personnel into glorified, unpaid "roadside assistance" providers.
The sentiment was echoed strongly by city leaders present at the hearing. Supervisor Bilal Mahmood directly challenged Waymo representatives on their strategy to prevent first responders from being diverted from their primary duties. When Sam Cooper, Waymo’s incident response specialist, pointed to the company’s efforts to train over 1,000 first responders in San Francisco through an online program, Mahmood was visibly unimpressed. He retorted, "Frankly, what I’m hearing mostly is that you kind of still expect our first responders to do roadside assistance, and you are just going to help us train them better to do that. I’m not really hearing a response about how you can take on some of that responsibility as well." This direct criticism underscored the city’s belief that Waymo is offloading operational burdens onto public services rather than internalizing them as part of its own service model. Supervisor Alan Wong further amplified the frustration, succinctly stating, "Our first responders should not be AAA roadside assistance."
This growing discontent in San Francisco is not an isolated incident but rather part of a broader pattern of disruptions caused by driverless vehicles. Just days before the hearing, a Waymo robotaxi gained notoriety in Austin, Texas, for blocking an ambulance attempting to respond to the scene of a mass shooting. Footage of the incident showed the autonomous cab inexplicably freezing in the middle of a street, forcing the ambulance driver to reverse and seek an alternative, potentially longer, route – a delay that could have critical consequences in a life-or-death situation. Such incidents are particularly egregious given the sensitive nature of emergency response. Earlier, Waymo cabs had been documented blazing through an active police standoff, seemingly oblivious to the volatile human drama unfolding around them. Other incidents include Waymo vehicles forcing police officers to pull them over for traffic infractions, such as driving on the wrong side of the road, further illustrating the unique and unexpected challenges posed by their unpredictable behavior. These events cumulatively paint a picture of a technology that, while promising, is still far from seamlessly integrated into the complex, dynamic environment of urban infrastructure and human-driven systems.
Waymo, for its part, acknowledges the need for collaboration with emergency services. A company spokesperson, in a statement to SFGATE, explained that the capability for emergency workers to disengage a Waymo vehicle’s driving system and manually move it was developed and implemented specifically at the request of first responders themselves. This feature, designed to allow responders to quickly clear a blocked path, was intended as a failsafe. However, the spokesperson emphasized, "We recognize the need for emergency responders to have this capability, but do not seek to make it the default. Waymo aims to limit the use of this feature as much as possible, prioritizing moving out of the lane of travel fully autonomously when appropriate and able." This statement, while attempting to reassure, does little to alleviate the core frustration of city officials who perceive that the "default" situation is precisely what is happening—responders are constantly called upon to manually intervene.
The underlying challenge lies in the current limitations of autonomous driving technology. While Waymo vehicles are equipped with an array of sophisticated sensors, cameras, radar, and lidar, enabling them to perceive their surroundings in intricate detail, they still struggle with truly novel or ambiguous situations. Power outages, complex accident scenes, or sudden, unpredicted human behavior can overload their decision-making algorithms, leading to a "safe but stalled" response. In these instances, the vehicles opt to stop entirely rather than risk a potentially dangerous maneuver, effectively trading collision risk for immobility. This precautionary programming, while understandable from a safety perspective, creates a logistical nightmare for urban centers that rely on efficient traffic flow, especially for emergency services.
The economic implications of this situation are also significant. When police, fire, or EMT units are diverted to manage a stalled robotaxi, taxpayer-funded resources are being expended. The cost of personnel, vehicle operation, and the opportunity cost of not being available for a real emergency adds up. Currently, there appears to be no mechanism for Waymo or other AV companies to reimburse cities for these "roadside assistance" services, effectively shifting a portion of their operational costs onto public funds. This financial burden, coupled with the safety risks, fuels the city’s demand for Waymo to take greater ownership of its vehicles’ failures.
The regulatory landscape further complicates matters. In California, the deployment and operation of autonomous vehicles are primarily overseen by the California Public Utilities Commission (CPUC), not directly by individual cities. This structure often leaves local municipalities feeling powerless to enforce their own regulations or demand specific operational changes from AV companies, even when public safety is at stake. The San Francisco public hearing, therefore, served not only as a platform for grievances but also as a political maneuver to exert pressure on Waymo and, indirectly, on the CPUC to reconsider the scope of AV operations and the responsibilities of the companies.
As autonomous vehicle technology continues its rapid advancement and expands its footprint in urban environments, the incidents in San Francisco, Austin, and elsewhere underscore a critical need for more robust, proactive collaboration between AV developers and municipal emergency services. Solutions could include Waymo developing more sophisticated protocols for handling unexpected disruptions—perhaps a dedicated, rapidly deployable response team that can reach stalled vehicles within minutes, or enhanced remote intervention capabilities that go beyond simple "disengagement." Furthermore, clearer communication channels with emergency dispatchers, real-time location sharing of stalled vehicles, and even geofencing technologies that could prevent AVs from entering critical incident zones might be necessary. The goal should be to minimize reliance on first responders for non-emergency tasks, allowing them to focus on their primary mission of protecting lives and property. Without such comprehensive solutions, the promise of driverless vehicles as a safe and efficient mode of transport risks being overshadowed by their disruptive potential and the growing frustration of the communities they serve.
