GM’s Cruise Rethinks Its Robotaxi Strategy After Admitting a Software Fault in Gruesome Crash

Cruise set out to win the autonomous car race by starting with urban driving. After a pedestrian was dragged under a robotaxi, the company and its parent GM are cutting jobs and making other changes.
Cruise autonomous car with its doors open and a row of other Cruise cars in the background
Cruise autonomous vehicles in Austin, Texas, on August 31, 2023.Photograph: Bob Daemmrich/Alamy

In August 2016, WIRED visited the San Francisco offices of a young startup recently snapped up by a surprising buyer. General Motors acquired three-year-old Cruise for a reported $1 billion in hopes the straitlaced Detroit automaker could coopt the self-driving technology tipped to disrupt the auto industry. Cruise CEO Kyle Vogt—a scrappy Twitch cofounder who competed as a teen in BattleBots—said he intended to stick around, but to keep running the driverless-car tech developer like a startup. He’d be out of a job, he predicted, if he couldn’t hack the self-driving thing in 10 to 15 years.

Is Vogt’s time up? GM’s financial reports show it losing $​​8.2 billion on Cruise since the start of 2017, and it has sunk at least $1.9 billion into the company this year. But last month, California regulators yanked its permits to operate self-driving vehicles in San Francisco, amidst allegations the company failed to disclose important details about a serious collision in which a pedestrian was trapped under a robot taxi. A few days later, the company said it would pause driverless operations across the US, in cities including Austin, Texas, and Phoenix, Arizona.

This week revealed new details of its technology’s failings during the San Francisco collision on October 2. On that night, a pedestrian was struck by a human-driven car and thrown into the path of a driverless Cruise vehicle that swerved but still hit the woman. Cruise said Wednesday that the car’s software then “inaccurately characterized” the collision as a side impact, not a front strike, and so automatically attempted to pull out of traffic, a maneuver that dragged her 20 feet along the ground. Cruise recalled all 950 driverless vehicles in its fleet, acknowledging that their software creates a safety risk, and says it will only resume driverless operations after updating it. (The person behind the wheel of the car that initially hit the woman has not been caught.)

GM now appears to have decided to tighten the leash on Cruise. As Forbes first reported Wednesday, layoffs have arrived. In an all-hands meeting Monday ​​focused on Cruise’s response to its trouble in California, CEO Vogt told employees that a timeline for job eliminations would come in the next few weeks. The company began laying off contract workers in cleaning, charging, and maintenance roles today. GM also said this week it would temporarily halt production of the Origin, a purpose-built robotaxi vehicle that Cruise had been testing in San Francisco and Austin.

“We believe strongly in Cruise’s mission and the transformative technology it is developing,” GM spokesperson Aimee Ridella said in a statement. “Safety has to be our top priority, and we fully support the actions that Cruise leadership is taking to ensure that it is putting safety first and building trust and credibility.”

Second Thoughts

Cruise’s initial response to the October crash suggested it was a freak incident—one unavoidable by even a human driver. Its car “responded to the individual deflected in its path within 460 milliseconds, faster than most human drivers, and braked aggressively to minimize the impact," the company said. This week’s recall and Cruise’s other recent actions seem to show the company conceding the possibility of systemic flaws in its strategy, technology, and communications with a nervous public.

Cruise said in a blog post Wednesday that it would increase transparency, and that it had retained a law firm to review the October crash and an independent engineering firm to review all of its safety and engineering processes. “As we build a better Cruise, we’re evaluating a variety of potential actions to ensure we operate at the highest standards of safety, transparency, and accountability,” Cruise spokesperson Navideh Forghani wrote in a statement.

And though the fallout from the San Francisco collision has led to Cruise’s most recent troubles, it’s becoming clear that the robotaxi operator faced pushback from other cities as well. Documents obtained by WIRED through a public records request from the city of Austin show that in the months before the company paused driverless operations at the end of last month, it had garnered complaints from the city’s fire, police, and emergency services departments, as well as residents—similar to criticism leveled by their counterparts in San Francisco.

A log maintained by the city of Austin notes at least 12 incidents between July and November characterized by city staff as “near misses.” In one, a firefighter reported he was almost hit by a Cruise vehicle that drove through the scene of a two-vehicle collision in front of downtown Austin’s Fire Station 2. Cruise says its car, guided by a remote assistance worker, followed human-driven cars and passed the firefighter at a distance of about 9 feet, but as in San Francisco, the Austin emergency departments appeared fed up—and concerned about potential disaster.

“We are having to[o] many issues here!” Austin Fire Department captain Matthew McElearney wrote in an October email to city officials and Cruise liaisons. He asked the company to set up an “avoidance area” around the fire station to ensure no driverless vehicles would pass by; a representative for the company said Cruise would do so as it investigated.

The logs also show the Austin Police Department complained in August and then October that the Cruise driverless vehicles did not understand hand signals or commands from traffic police. “Biggest and probably the most dangerous issues [sic] I’ve seen with them is when we are directing traffic,” one police official wrote, noting that if police issued commands contrary to traffic lights, the cars “will blow through or just stop.” Incidents occurred “almost nightly” around the University of Texas at Austin’s Moody Center arena, the log said.

In a written statement, Cruise spokesperson Forghani said, “We value the work of our police officers and first responders and maintain an open line of communication with them to discuss their concerns.” In Austin, the company meets monthly with a task force that includes several city department representatives to discuss issues.

Competitive Drive

Cruise was founded years after the Google self-driving project that became Waymo began, but it put riskier, urban driving at the center of its mission. Vogt has insisted that that strategy will be key to winning the driverless race. Waymo, now Cruise’s chief competitor, by contrast, started its initial Uber-like service in the suburbs of Phoenix, while Cruise launched driverless testing and service in more complex San Francisco.

Cruise persisted in launching its service in the city despite reports that its vehicles sometimes froze in the middle of the traffic, impeded emergency responders on their way to fires and crashes, and held up city buses and streetcars. Over the summer, Cruise (and Waymo) won permits to carry paid San Francisco passengers in its driverless cars throughout the day, but not before a marathon hearing in which fire and police officials, as well as representatives from local taxi and driver groups, waited for hours to air their complaints about the cars.

GM seemed eager to give Vogt and his once-startup the freedom to be aggressive in rolling out robotaxis. Leadership put the milder-mannered company veteran Dan Ammann in charge of the subsidiary in 2018, then recalled him and reinstalled Vogt in 2021 after the company failed to launch driverless service by a self-imposed deadline. In 2022, GM committed itself further to Vogt’s approach when it bought out Softbank’s stake in Cruise for another $3.4 billion.

That investment and the other billions GM has plowed into Cruise now look a little questionable. For industry watchers, the situation feels familiar. In 2018, a self-driving vehicle being tested by Uber struck and killed a woman in Arizona. The company had gained worldwide notoriety for a “move fast and break things” approach to transportation.

After the crash, Uber completely reorganized its safety procedures, as Cruise suggested it might do this week. Still, Uber sold off its self-driving unit two years later. On Tuesday, the day before Cruise’s recall, Uber’s current CEO, Dara Khosrowshahi, became an unlikely cheerleader for prudence in an interview on CNBC. “You can’t take shortcuts as it relates to safety,” he said when asked about the future of driverless technology. “So I do think that the regulatory framework, the safety framework, still has yet to mature, and I do think this is something all of us should take our time on.”

Kyle Vogt’s impulses appear to retain a Silicon Valley ethos. This weekend, he logged onto the message board Hacker News, operated by legendary startup incubator Ycombinator, to clarify a statistic that appeared in a blistering New York Times story that alleged culture and tech problems at the self-driving developer. (Cruise had rebuffed the Times’ request to speak to Vogt before publication.) Meanwhile, his robotaxi company still isn’t running service, and GM is is spending more than $500 million every quarter to operate it. That clock Vogt mentioned back in 2016 keeps ticking.