Welcome back to SavvyMonk, your one-stop for AI and tech news that actually matters.

This week, over 100 driverless robotaxis simultaneously stopped moving in the middle of busy roads in Wuhan, China. Passengers were trapped inside. Police had to respond. And Baidu, the company behind it all, went very quiet.

Let's get into it.

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

Just before 9 PM on Tuesday, dozens of Baidu Apollo Go robotaxis stopped moving across Wuhan, and they didn't pull over gracefully or slow to a safe stop. They simply froze wherever they were, mid-lane, on busy roads and elevated ring highways, with traffic streaming past on both sides.

Videos spread quickly on Weibo, showing vehicles stationary at intersections with hazard lights blinking, and one clip appeared to show a following car colliding with a frozen robotaxi on a highway, though police later confirmed no injuries resulted from the crash.

Passengers inside called customer service and found the line overwhelmed, and one rider documented a 90-minute ordeal stuck on an overpass surrounded by dump trucks, with her order eventually cancelled while she was still sitting inside the car. Some passengers were too afraid to exit at all, because the vehicles had stopped in positions where stepping out meant stepping into fast-moving traffic, so police were called to assist them instead.

By the time the situation was resolved, authorities confirmed that at least 100 Apollo Go vehicles were affected in Wuhan, and broader reports suggested the outage extended across multiple cities including Beijing and Shenzhen, with the total fleet impact closer to 1,200 cars.

What Caused This

Baidu attributed the failure to a "technical issue in our backend dispatch system," and has not shared anything more specific beyond that. Wuhan traffic police said preliminary findings pointed to a system malfunction and noted the investigation was still ongoing.

That framing matters more than it might seem, because the failure was not a single vehicle's sensors going wrong or one car misjudging a turn. When over 100 vehicles stop at exactly the same time and in exactly the same way, the problem is upstream, almost certainly in the centralized cloud-based fleet management systems that every car in the network depends on.

Analysts are already speculating whether the outage originated in Baidu's own cloud division or in third-party infrastructure it relies on, but either way the result is what experts call a correlated failure, where one problem takes down everything simultaneously rather than staying isolated to a single vehicle.

Why Wuhan

Wuhan is not a test city for Baidu, and it is important to understand just how far along this deployment actually was.

The city is home to Baidu's largest Apollo Go fleet, with over 1,000 fully driverless vehicles operating commercially across its roads, meaning there is no safety driver in the back seat and no remote copilot physically in the car.

In the fourth quarter of 2025 alone, Apollo Go delivered 3.4 million fully driverless rides, with weekly rides topping 300,000 during peak periods, and the company had publicly positioned Wuhan as a global showcase for what mature autonomous mobility looks like at scale.

Which is exactly what makes Tuesday's incident something qualitatively different from a glitch in a pilot programme.

The Bigger Picture

This is not Baidu's first bad headline, and the pattern of incidents is worth knowing before drawing conclusions about Tuesday alone. In December 2025, authorities in Zhuzhou suspended robotaxi operations after a Baidu vehicle struck two pedestrians, leaving them in intensive care.

In August 2025, an Apollo Go car in Chongqing drove into a construction pit with a passenger inside. And the problem is not unique to China either, because in December a power outage in San Francisco caused Waymo's fleet to stall and clog traffic across the city, pointing to a shared vulnerability across the entire category of autonomous fleet operations.

But the scale of Tuesday's failure is still different from anything that has come before it, because no robotaxi operator has ever had this many vehicles go offline simultaneously during active commercial service in a major city. It is, in a real sense, the first genuine stress test of what happens when a city-scale autonomous fleet hits a backend catastrophe, and the results were not encouraging for anyone watching.

The timing is also particularly uncomfortable for Baidu's global ambitions, because in December 2025, Uber and Lyft both announced partnerships with the company to bring Apollo Go vehicles to London.

Uber's pilot was expected to begin in the first half of 2026 using the Apollo Go RT6, a purpose-built electric vehicle with a detachable steering wheel, and Lyft's rollout was set to start with 50 vehicles following its $197 million acquisition of the FreeNow mobility platform.

Neither trial has regulatory approval yet, and footage of frozen robotaxis blocking highways in China is unlikely to make those conversations move any faster.

Baidu's Hong Kong-listed shares dropped 6.2% on Tuesday, while Pony.ai fell 3.1% and WeRide dropped 2.8%, even though neither of those companies experienced any outage of their own.

What This Means for Regulation

The regulatory frameworks that govern vehicles were built around individual failures, things like a sensor error, a driver error, or a mechanical defect in a single car, and they were not designed for a scenario where 100 vehicles fail the same way at the same moment across an entire city.

Jack Stilgoe, a professor of science and technology policy at University College London, told BBC News that while driverless technology may be safer on average than human drivers, this incident showed it could "still go wrong in completely new ways," and that distinction is the one that policymakers are going to need to sit with.

The question is no longer whether autonomous vehicles can drive safely most of the time, because the data on that is actually quite good. The question is whether the systems that manage, monitor, and recover them are resilient enough for what happens when they don't.

The Bottom Line

Baidu's Wuhan outage is not evidence that robotaxis are inherently unsafe, and the cumulative safety record across millions of rides is genuinely strong.

But this incident introduced a new category of failure that the industry and its regulators have not had to seriously reckon with before, because a single car crash is a tragedy while a hundred cars stopping at once is a systems problem, and those two things require completely different responses.

Right now, nobody, including Baidu, has a clear public answer for how you prevent it from happening again.

Category: Crisis Communication

"Write a public statement for [Company Name] in response to a product failure that affected [number] customers. The tone should be direct and honest without being defensive. Acknowledge what happened in plain language, explain what is currently known about the cause, describe the immediate steps being taken, and commit to a specific follow-up date for a fuller update. Avoid corporate jargon and do not speculate about the cause if it is still under investigation."

The strange thing about watching 100 robot cars freeze simultaneously is that there was no one to call, no driver, no dispatcher, no person in the loop who could simply pull over and handle it.

Baidu's customer service line was overwhelmed, passengers sat on highway overpasses in the dark unable to exit safely, and the technology worked perfectly right up until it didn't, at which point there was no fallback at all.

That gap between "mostly works" and "safe when it fails" is exactly where regulators, operators, and the public need to start asking much harder questions. Hit reply, I read every response.

See you in the next one.

— Vivek

P.S. If you know someone who wants AI and tech news without the hype, send this their way. They can subscribe at https://savvymonk.beehiiv.com/