Alphabet Inc.-owned Waymo has initiated a voluntary recall of approximately 3,800 robotaxis following a harrowing incident in San Antonio where an autonomous vehicle was swept into a creek.
The recall, detailed in documents posted Tuesday by the National Highway Traffic Safety Administration (NHTSA), underscores the ongoing technical hurdles facing the driverless car industry during extreme weather events.
The federal filing reveals that the recall affects vehicles equipped with Waymo’s fifth- and sixth-generation automated driving systems (ADS). The safety probe was triggered by an April 20 incident in which an unoccupied Waymo vehicle traveling at 40 mph entered a flooded roadway in San Antonio. Despite detecting the water, the system only reduced its speed rather than coming to a full stop, resulting in the vehicle being washed away.
Waymo officials said they have “identified an area of improvement regarding untraversable flooded lanes specific to higher-speed roadways.” While the company’s Waymo Driver system utilizes a sophisticated suite of LiDAR, radar, and cameras to act as a “mobile weather station,” the Texas incident exposed a critical software gap: the inability of the ADS to correctly prioritize flood depth over maintainable speeds on fast-moving roads.
The recall comes at a sensitive time for the company, which is aggressively expanding its commercial footprint. Recently, Waymo has faced mounting scrutiny for several high-profile malfunctions, including:
Gridlock in San Francisco. Vehicles stalled during widespread power outages in December, blocking emergency routes.
Safety violations in Austin, Texas. Reports of robotaxis failing to yield to school buses.
Stalling in floods. Footage from Austin recently captured multiple Waymo vehicles stalling in rising waters, forcing human drivers to navigate around the high-tech obstacles.
In response to the latest failure, Waymo has temporarily suspended service in San Antonio. The company confirmed it has already deployed software updates across its fleet and implemented new “operational restrictions” to prevent robotaxis from entering known flash-flood zones during intense rainfall.
“Safety is our primary priority,” the company said in a statement, noting that it provides more than 500,000 trips weekly. Waymo is also reportedly collaborating with Google DeepMind to utilize advanced simulation tools, stress-testing the AI against “tornado-like conditions” and other rare “edge cases” that are difficult to replicate in real-world testing.
The recall also highlights a stark reality for passengers. In the event of a weather emergency, riders have no direct control over the vehicle’s driving behavior. While passengers can request to “pull over” via an in-car screen or mobile app, the AI ultimately decides where and when it is legally and physically safe to stop. If the system deems conditions too hazardous to move, passengers cannot force the vehicle to continue, leaving them reliant on the software’s internal risk assessment.
Affected vehicles include those produced between March 2022 and April 2026. Waymo says it is currently “readying operations to resume” in San Antonio as it continues to monitor the performance of its updated software safeguards.
Despite the hiccups and recall, Waymo remains a far more secure option to human drivers, argue analysts, who point to the brand’s safety record. “Autonomous vehicles are held to an unrealistic standard. They have to be perfect when humans are absolutely not,” Steven Dickens, CEO of HyperFRAME Research, said. “The strive for perfection hampers adoption and an overall safety improvement on our roads. A Waymo being safe (they are) is graded on a different scale to humans and that is a shame.”
The voluntary recall of 3,800 Waymo robotaxis “highlights a critical gap in the industry’s ability to navigate edge-case environmental hazards such as flash flooding,” added Ron Westfall, another HyperFRAME Research analyst.
“By identifying that the ADS only decelerated rather than halting completely, the NHTSA filing reveals a specific failure in the system’s object and event detection logic when faced with non-solid obstacles,” Westfall said. “This incident across two generations of automated driving systems indicates that even advanced hardware requires more robust software training to distinguish between traversable puddles and life-threatening water currents.
“This also provides Waymo’s competitors, such as Tesla or Zoox, with a lessons learned blueprint to refine their own weather-detection algorithms and safety protocols without incurring the same reputational or regulatory costs. I see this recall serving as a sobering reminder that achieving full autonomy requires solving not just high-traffic urban scenarios, but also the unpredictable physics of extreme weather.”
