Robotaxis hit the same wall in two cities: flooded roads and construction zones

Waymo has now halted service in Atlanta, San Antonio, Dallas and Houston, TechCrunch reported on 21 May 2026. The trigger was a flooded street in Atlanta that Wednesday. An unoccupied vehicle stopped there and stayed put for roughly an hour before crews recovered it.

The company framed the pause as weather caution. "Safety is Waymo's top priority, both for our riders and everyone we share the road with. During a period of intense rain yesterday in Atlanta, an unoccupied Waymo vehicle encountered a flooded road and stopped," Waymo said in a statement given to TechCrunch. A spokesperson said the company paused Dallas and Houston out of an abundance of caution over forecast severe weather.

Timing is what makes this matter. Waymo had already issued a software recall the week before over the same class of problem. In documents released by the National Highway Traffic Safety Administration, the company admitted it had not finished a "final remedy" for avoiding flooded areas. What it shipped instead were "restrictions at times and in locations where there is an elevated risk of encountering a flooded, higher-speed roadway."

One month, two different failures

The water problem is not the only open front. On 18 June 2026, Waymo recalled nearly 4,000 robotaxis to keep them off highways while it works out how they should behave around construction zones. TechCrunch reported that Waymo had identified at least 13 instances of its vehicles driving into highway sections closed for construction: six in Phoenix in April, seven in San Francisco in May.

Waymo pulled its robotaxis from all highways on 19 May, and the fix is "currently under development," per NHTSA filings. The company is not pulling vehicles off the road entirely. Surface-street service continues, though it has periodically paused during severe weather. In its own statement, Waymo said: "We identified an area of improvement regarding performance around freeway construction zones. We voluntarily restricted freeway operations last month while making improvements, proactively notified state and federal regulators, and decided to file a voluntary software recall with NHTSA."

By TechCrunch's count, this is the sixth recall Waymo has issued for its robotaxis. The list includes a December recall over illegal behavior around school buses, plus earlier fixes for low-speed collisions with chains, gates and telephone poles, and a problem involving towed trucks. The school-bus behavior is now the subject of active investigations by both NHTSA and the National Transportation Safety Board. A second set of investigations covers a 23 January incident in Santa Monica, California, where a Waymo robotaxi struck a child. Waymo has said the vehicle braked to around six miles per hour before impact and that the child suffered minor injuries.

"We identified an area of improvement regarding performance around freeway construction zones," Waymo said in a statement to TechCrunch.

The scale of the operation makes each edge case more visible. Waymo says its vehicles have driven more than 170 million miles autonomously and claims a 13x reduction in serious-injury-or-worse crashes compared with human drivers. It is also planning to launch in more than 20 cities this year alone, including London and Tokyo. Highway rides only started in November 2025.

The Tesla comparison is unavoidable, and awkward

Tesla's robotaxi launch in Austin on 22 June 2025 offers a different failure profile. Business Insider asked three autonomous driving experts to review rider videos from the early-access period, when rides were limited to influencers and investors. The issues those experts flagged included speeding more than 10 miles above the limit, driving into the wrong lane, random braking, pullover button problems, and at least two safety-monitor interventions over parking.

Tesla's system relies on an advanced version of Full Self-Driving with eight cameras and no radar or lidar, unlike Waymo. Raj Rajkumar, a professor of engineering at Carnegie Mellon University, told Business Insider that pullover and wrong-lane errors could likely be fixed with more training data, but that so-called phantom braking may point to a design flaw. "To process camera data, one has to use AI and machine learning," he said. "But hallucinations are an integral part of how AI operates, and once you hallucinate, phantom braking ends up happening, so a camera-only solution will not be sufficient for a very long time."

Steven Shladover, lead researcher at the Partners for Advanced Transportation Technology program at UC Berkeley, told Business Insider that automated driving needs cameras, radar and lidar plus precise localization against high-accuracy maps. He said he worries Tesla's camera-only approach will eventually lead to passenger injuries without intervention.

Phantom braking is not new to Tesla. The company faces a class-action lawsuit over alleged phantom braking in Autopilot, a driver-assistance system distinct from supervised Full Self-Driving. NHTSA opened an investigation in 2022 into Model 3 and Model Y vehicles after more than 750 drivers complained of sudden high-speed braking on Autopilot. Business Insider reported that investigation and evaluation are still ongoing.

Separately, an April 2026 RTS investigation based on leaked internal Tesla data reported that the company had concealed thousands of serious incidents tied to its autonomous driving systems through 2022, some fatal. The report said more than 2,400 complaints involved spontaneous acceleration and that accidents exceeded 1,000, with many marked "unresolved." It also described a jury verdict ordering Tesla to pay more than $243 million in damages in a case involving a pedestrian killed by a Tesla in Autopilot mode, and said a federal judge upheld the award in late February. Tesla has said it can still appeal.

Two companies, two sets of edge cases, one shared lesson for anyone watching the sector: the hard part is not the demo. It is the flooded intersection, the coned-off ramp and the phantom object that only shows up in the training data after someone has already hit it.