Waymo’s autonomous robotaxis are supposed to be ready for the real world. They’re not. The company has now paused service in four cities because its vehicles keep getting stuck in flooded streets, a problem so persistent that it prompted a software recall last week. If a self-driving car can’t handle rain, we have to ask: what exactly is it ready for?
The latest incident happened in Atlanta on Wednesday when one of Waymo’s robotaxis drove into a flooded intersection and got stranded for about an hour before being recovered. This wasn’t some freak accident. The flooding was so sudden that it actually outpaced the National Weather Service alerts that Waymo relies on to prepare its fleet. In other words, the weather happened faster than the warnings could warn about it.
That’s the kind of edge case that’s supposed to be easy for autonomous systems to handle. Rain exists everywhere. Flooding happens regularly in cities. Yet here we are, watching Waymo pull service from San Antonio, Texas, Dallas, Houston, and now Atlanta because its vehicles can’t navigate wet roads safely.
A Fix That Didn’t Actually Fix Anything
Last week, Waymo admitted something telling: it hadn’t actually finished developing a “final remedy” for avoiding flooded areas. Instead, the company shipped what it called “restrictions at times and in locations where there is an elevated risk of encountering a flooded, higher-speed roadway.” Basically, the company told its cars to be more careful in certain areas during certain times. That’s not a solution. That’s a band-aid wrapped around admitting the problem wasn’t solved.
And the band-aid failed. The Atlanta robotaxi encountered that flooded intersection despite these supposedly precautionary measures. Waymo later explained that the storm came so fast that the vehicles didn’t have time to react based on weather alerts, but that explanation only raises more questions. If your autonomous vehicle system is that dependent on external alerts, how autonomous is it really?
The company told TechCrunch that “safety is our top priority,” a statement that feels increasingly hollow each time something goes wrong. Safety being a priority and safety being actually achieved are two different things.
A Pattern of Playing Catch-Up
This isn’t Waymo’s first rodeo with having to chase problems after deploying fixes. Last year, people noticed Waymo robotaxis were illegally passing stopped school buses. The company shipped a software update to fix it. The vehicles kept doing it anyway. Now both the National Highway Traffic Safety Administration (NHTSA) and the National Transportation Safety Board (NTSB) are investigating that issue. The NHTSA even sent Waymo a second document request because the company’s initial response didn’t provide enough data.
There’s also an active investigation into a January incident in Santa Monica where a Waymo robotaxi hit a child. Waymo said it was braking to six miles per hour at the time and the child suffered minor injuries, but the fact that this situation even required investigation speaks to a broader pattern.
The company is reactive, not proactive. It deploys fixes after problems surface instead of catching them beforehand. For an autonomous vehicle fleet operating in public streets, that’s not a great look. It’s also not a sustainable way to build trust.
What This Means for the Broader Push
Waymo isn’t the only company betting on autonomous vehicles, but it’s one of the most visible. When Waymo struggles with weather and school buses, it’s not just a Waymo problem. It’s a technology problem, one that suggests the autonomous vehicle industry might be moving faster than its actual capabilities.
The regulatory scrutiny is only going to intensify. NHTSA isn’t sitting on the sidelines anymore. It’s actively requesting documents, asking follow-up questions, and watching closely. That’s what happens when a company’s public-facing product keeps failing in ways that affect safety.
Waymo’s safety challenges with rain, school buses, and pedestrians aren’t edge cases. They’re core use cases. And if a company can’t handle them consistently, then the question isn’t whether autonomous vehicles will eventually work. The question is whether we’re willing to keep paying the price while they learn.
