Mass robotaxi malfunction halts traffic in Chinese city
On the evening of March 31, 2026, more than 100 Baidu Apollo Go robotaxis stopped moving simultaneously on roads across Wuhan, China. They did not pull over. They did not execute a graceful shutdown. They simply froze, scattering across elevated ring roads and intersections while traffic streamed past on both sides. Passengers trapped inside called the police. Some were stranded for nearly two hours before help arrived, Reuters reported.
Wuhan traffic police said in a statement that initial findings pointed to a system malfunction. The cause is under investigation. Baidu, which owns and operates Apollo Go, has not issued an official public statement. Operations had resumed as of Tuesday, according to CnEVPost.
The city where this happened is not a test track. Wuhan is home to Apollo Go's largest fleet in China, with more than 1,000 driverless vehicles operating across its roads. This is not a pilot program. In the fourth quarter of 2025 alone, the service delivered 3.4 million fully driverless rides, with weekly rides surpassing 300,000 during peak weeks, Baidu said in its February 2026 earnings report. By that date, Apollo Go had completed more than 20 million rides cumulatively across 26 cities globally. The scale makes this qualitatively different from an isolated mechanical failure. When 100 human drivers have car trouble, 100 individual incidents happen. When 100 robotaxis fail simultaneously from a shared system trigger, that is a different risk category entirely.
Industry insiders speaking to The Paper, a Shanghai-based news outlet, attributed the Wuhan paralysis to a safety self-check mechanism triggered unexpectedly. Insiders described this as a proactive safety strategy, though the result left passengers stranded on elevated highways in heavy traffic. The framing is revealing: the same safety logic that prevents a robotaxi from driving into a barrier is the logic that can freeze an entire fleet at once.
Jack Stilgoe, a professor of science and technology policy at University College London, told BBC News that while autonomous driving may be safer on average than human drivers, this incident demonstrated it could still fail in completely new ways. "A human driver who has a medical episode creates one hazard," Stilgoe said. "A fleet-wide failure creates dozens simultaneously. If we are going to make good choices about this technology, we need to understand entirely new types of risk."
The Wuhan incident lands uncomfortably close to an Apollo Go expansion milestone. On March 30, Baidu announced the launch of fully driverless commercial operations in Dubai via the Apollo Go app. The expansion is the context that makes Tuesday's failure awkward in timing.
This is not the first time a robotaxi fleet has stalled simultaneously. In December 2025, a power outage at a Pacific Gas and Electric substation knocked out electricity across roughly a third of San Francisco, disabling traffic signals across multiple neighborhoods. Waymo's vehicles, which treat dead traffic signals as four-way stops, began requesting confirmation checks from the company's human fleet response team. The concentrated spike in requests overwhelmed the system. Dozens of vehicles stopped in place, blocking streets and contributing to significant traffic jams, The Next Web reported. Waymo suspended service and later shipped a software update to help vehicles navigate disabled traffic infrastructure more decisively.
The parallel is instructive: the San Francisco incident was triggered by external infrastructure failure, while the Wuhan failure appears to originate from within the autonomous system itself. But both ended with the same operational outcome, and both raise the same question. When a robotaxi encounters a situation it cannot resolve safely, it enters what the industry calls "minimal risk condition" — essentially, it stops and waits for help. That protocol makes sense for one vehicle. It creates a traffic obstruction when dozens of vehicles execute it simultaneously on city streets.
China's robotaxi industry has moved faster than its Western counterparts in deploying at scale, in part because regulators there have been willing to grant commercial operating permits more quickly. Baidu, Pony.ai, and WeRide have all expanded commercial services across major Chinese cities and are now pushing into overseas markets. Apollo Go's Dubai launch this week is the most recent example. As these fleets grow, the failure surface grows with them. A system that works reliably at 500 vehicles may fail in ways that did not exist at 500 vehicles when it operates 1,000 or 10,000. That is not an argument against the technology. It is an argument for taking seriously what a mass fleet failure actually looks like, rather than treating it as a theoretical edge case.
The China Insurance Industry Association is reportedly finalizing insurance terms specifically for driverless vehicles, according to Yicai Global. That work was already underway before Tuesday's incident. It may need to move faster now.
No injuries were reported in Wuhan. All passengers exited their vehicles safely, police said. That is the most important fact in this story, and it is worth holding onto. The question is what happens the next time 100 vehicles freeze on a city street, and the next, and the one after that.