When transit systems experience delays, the reason usually isn’t very interesting: congested streets, medical emergencies, mechanical problems. But the cause of a recent holdup on San Francisco’s MUNI system at least had the virtue of being novel.
On Sept. 30 at around 11 p.m., an N Line streetcar ground to a halt at the intersection of Carl Street and Cole Street because an autonomous vehicle from Cruise, a subsidiary of General Motors, had halted on the streetcar tracks and wouldn’t budge. According to the city’s transportation department, the 140 passengers riding the N line that evening were stuck in place for seven minutes before a Cruise employee arrived and moved the driverless conveyance. (Cruise did not respond to questions about what happened that night.)
This incident, which was not reported in the media at the time, is one of many in which autonomous vehicles roaming San Francisco’s streets have disrupted the city’s transportation network. In April, a Cruise vehicle blocked a travel lane needed by a siren-blaring fire engine, delaying its arrival at a three-alarm fire. Last fall, dozens of self-driving cars from Google’s Waymo subsidiary drove daily into a quiet cul-de-sac before turning around, much to the frustration of nearby residents.
Because of California’s insufficient and outdated AV reporting requirements, many incidents like these have escaped both public attention and regulatory consequences. Facing minimal scrutiny, AV companies have little incentive to avoid mucking up the public right of way—or even keep city officials informed about what’s happening on their streets.
With Silicon Valley a few miles away, San Francisco has become the top urban location for AV testing and deployment. With California officials granting their first AV deployment permits allowing passenger service this year, the city now offers a preview of what’s to come in other places where self-driving companies are now fanning out, with expansions announced for Los Angeles, Las Vegas, Phoenix, and Austin.
Based on San Francisco’s experience, residents and officials in those cities should brace for strange, disruptive, and dangerous happenings on their streets. And they should demand that state officials offer the protection that California is failing to provide.
After a decade of testing and hype, it suddenly felt this year as if self-driving cars were everywhere in San Francisco. While individuals—in California or anywhere else—can’t buy these vehicles for themselves, companies are competing to improve the technology and roll out taxi services that resemble ride hail. In San Francisco, Cruise and Waymo allow residents to request a ride on their app, summoning a driverless vehicle that brings them to their destination. (Waymo transports passengers using safety drivers who can intervene if something goes wrong; Cruise does not.) Joining Cruise and Waymo robotaxis on San Francisco streets are testing vehicles from a number of other AV companies.
Two California agencies decide which autonomous companies have permission to operate within the state. The Department of Motor Vehicles issues permits for the vehicle itself (dozens of companies have obtained one), while the California Public Utilities Commission provides permits for passenger service. Earlier this year the CPUC issued Cruise and Waymo the state’s first robotaxi permits to transport paying passengers.
California requires that companies conducting AV testing submit information about collisions as well as “disengagements,” or moments when the autonomous system is forced to transfer driving responsibility to a human. The DMV publishes this information, along with each company’s total number of miles of autonomous driving on state roads. But AV executives—joined by some outside observers—have criticized disengagements as a deceptive metric, since it does not take into account the higher degree of difficulty navigating urban streets compared with interstates. Companies could also tinker with their disengagement data to seem safer than they are, something that the Chinese company AutoX has been accused of doing.
“It’s an open question whether or not disengagement data gives you anything useful,” said Billy Riggs, a professor at the University of San Francisco’s School of Management who has followed the state’s AV deployments closely. Even so, California’s regulatory focus on disengagements has powerfully shaped national media coverage of AV safety.
Strange as it may seem, California stops requiring that AV companies share disengagement data and collision locations as soon as they begin collecting passenger fares, as Waymo and Cruise now do. From that point forward, if an AV vehicle jeopardizes safety on the street—for instance, by causing a crash or blocking a transit line—the public won’t know unless the AV company chooses to publicize it (unlikely) or if a passerby reports the incident to 911 or posts about it on social media (unreliable).
Relaxing oversight for AVs with paying passengers might have seemed appropriate during the industry’s age of optimism several years ago, when policymakers could assume that companies would have “solved” autonomous driving before they started charging people for trips. If so, San Francisco’s experience shows that the reality is something else entirely.
Cities, for their part, play no defined role in the state’s AV regulatory structure, leaving them struggling to obtain information about AV-induced roadway blockages or even a list of companies deploying testing vehicles on their streets. Asked by email under what specific conditions Cruise notifies local leaders about an incident, spokesperson Hannah Lindow replied only that the company “maintains an open line of communication and meets regularly with city officials.”
Urban leaders anticipated these problems and tried to forestall them. In 2020 officials from Los Angeles, San Francisco, and San Diego asked the CPUC “not [to] create a deployment program that would give participants blanket authority to operate a fared service anywhere in the State.” The CPUC rejected that request, which has hobbled cities’ ability to manage their streets.
There is an ominous precedent for this situation. When ride hail emerged a decade ago, Uber and Lyft lobbied, mostly successfully, for states rather than cities to oversee them. That preemption left urban leaders with few tools to control (or even monitor) ride hail, which researchers have found increases traffic congestion and reduces transit ridership. Early evidence suggests that an influx of AVs could create similar problems, but on a much larger scale.
It’s been less than a year since California granted its first AV deployment permits, but robotaxis have already caused a slew of problems in San Francisco. In April, a Cruise vehicle stopped by city police pulled over to the side of the street—and then promptly drove away from an officer who tried to look inside. (“Are you serious? How does that happen?” a baffled onlooker exclaimed.) In June, a phalanx of at least a dozen Cruise vehicles obstructed a city arterial. (“Oh no, they’re plotting,” someone quipped on Reddit.) According to city officials, 28 incidents involving Cruise were reported to 911 between May 29 and Sept. 5, including instances of vehicles driving on the sidewalk.
Cruise and other AV companies maintain a ”Critical Response Line” dedicated to handling emergencies, but no public data measure their responsiveness. Cruise does not include information about right-of-way blockages or response times in its self-written, 175-page “safety report.” Asked how quickly Cruise responds to an emergency involving one of its vehicles, Lindow, the company spokesperson, said Cruise is “striving to do so in 10 minutes or less.” She did not reply when asked how often it achieves that benchmark.
Cruise seems to cause the lion’s share of San Francisco’s AV headaches, but other companies have created problems too, such as the Waymo vehicles that constantly drove into the dead-end terminus of 15th Avenue, waking up the neighborhood.
Despite all the issues in San Francisco, California’s regulatory agencies have shown no signs of tightening or revising their oversight. Mark Rosekind, the former head of the National Highway Traffic Safety Administration, or NHTSA, who now helms the nonprofit California Mobility Center, said authorities “should evolve their regulatory approach to reflect the current state of technology. What data are needed to effectively identify safety issues? That’s a conversation state officials should be having with AV companies and other stakeholders.”
In September, two San Francisco transportation departments took the unusual step of bringing their concerns about AV deployments—and Cruise in particular—directly to the federal government in a 39-page letter submitted to the NHTSA. The letter was prompted by General Motors’ request that NHTSA exempt Cruise’s new AV vehicle, the Origin, from federal vehicle safety rules. Although the city officials stated they neither “support nor oppose” GM’s request, recent experiences have clearly given them pause. The letter highlighted Cruise’s patchy response to emergencies: “On one occasion on August 4, 2022, a City dispatcher placed four calls over six minutes [to Cruise’s emergency response line]; none of these calls were picked up.” Lindow did not respond when asked for comment on that allegation.
The National Association of City Transportation Officials, representing municipal transportation departments across North America, submitted its own letter to the NHTSA that flatly opposed GM’s request that the Cruise Origin receive an exemption from vehicle safety rules. (NHTSA has not yet made a decision.) Kate Fillin-Yeh, NACTO’s director of strategy, said urban transportation leaders nationwide are watching events unfold in San Francisco with growing concern. “I know that AV companies can make more money in cities because there is a density of people there,” she said, “but they’re unhelpful to the many people who rely on transit or walk.”
Indeed, beyond the wow factor of stepping inside a self-driving car, it’s unclear how exactly the introduction of robotaxis improves an urban transportation network. But the risks—including disruptions on public roadways, increased congestion, and reduced transit use—are very real.
Fillin-Yeh said her top request for federal and state policymakers is that they empower local leaders to monitor and manage AVs using their streets. “Cities need to be a part of these conversations about permitting and regulating AVs,” she said. “That isn’t always happening.”
In their letter to NHTSA, San Francisco officials proposed several ways to improve AV oversight. They suggested that NHTSA treat “travel lane failures that block roadways” as a key measure of AV readiness, adding that NHTSA should also quantify and publicize AV companies’ response times to vehicle emergencies.
Riggs, the University of San Francisco professor, agreed on the need to evaluate AV companies’ emergency response times, adding that governments must be especially careful to protect so-called vulnerable road users. “We should be collecting autonomous vehicles’ near-misses with pedestrians and cyclists,” he said.
In an interview, San Francisco Municipal Transportation Agency director Jeffrey Tumlin insisted that his city’s goals do not conflict with those of AV companies. “It’s in the interest of both the city and the AV industry to minimize impacts to transit and emergency response time,” he said. “But the AV industry doesn’t work if all the vehicles are stuck in traffic congestion. And the industry doesn’t work if they lose all political support because they’re blocking transit and fire departments.”
Meanwhile, AV companies are turning their attention far beyond the Bay Area. Recent press releases have announced new AV deployments in cities across the West and Southwest. Waymo recently declared that it will begin operating in Los Angeles, and a Lyft/Motional partnership plans to launch there as well. (It already operates in Las Vegas.) Cruise, meanwhile, has opened a waitlist for new service in Austin, Texas, and Phoenix, and chief operating officer Gil West told Reuters, “You’ll likely see us expand the number of markets in a large number [in 2023].” (Southwestern states like Nevada, Texas, and Arizona generally have more laissez-faire approaches toward AV regulation than California does.)
Problems resembling those in San Francisco are already surfacing elsewhere. In November Waymo announced that its robotaxi service was available in downtown Phoenix; less than 10 days later, a passenger tweeted that her autonomous ride there “was smooth sailing until it got stuck in the middle of an intersection.”
With the demise of Argo.ai, an AV company that had received billions in investment, remaining companies face growing pressure to showcase their deployment capabilities to antsy investors. Such competition can be healthy—if it doesn’t sacrifice societal goals around safety, equity, and a balanced transportation network.
But the experience of San Francisco suggests that it very well could, which would undermine support for AVs writ large. After all, it’s difficult to see why Americans should embrace robotaxi services that block intersections, delay transit service, and slow emergency response times. And it’s even harder to understand why any state would repeat California’s mistake of allowing AV companies to provide robotaxi service without collecting and sharing information about all crashes, roadway obstructions, and incident response times. At the federal level, Congress’ interest in restricting states’ few AV management tools would be a step in precisely the wrong direction.
Unfortunately, that now seems to be where autonomous-vehicle regulation is: stuck in an intersection, like that Waymo car in Phoenix, with no one in the driver’s seat.
Lucas Peilert provided research assistance.