Waymo recalled approximately 3,800 robotaxis on May 12, 2026 [1], after an unoccupied vehicle was swept away by floodwaters in San Antonio, Texas [2].
This recall highlights a critical gap in the ability of autonomous driving systems to perceive and react to extreme environmental hazards. The incident suggests that current sensor suites and software may struggle to differentiate between drivable roads and dangerous water levels during flash floods.
The recall follows an incident where one robotaxi [3] entered a flooded road and lost control, eventually being washed away. Waymo said a risk of loss of vehicle control on flooded roads was the primary reason for the action [4]. The company said that the autonomous system could not reliably judge flooded road conditions, which led to the vehicle entering the water [5].
The affected fleet consists of nearly 3,800 vehicles [1]. This measure is intended to prevent similar occurrences across the company's wider operational areas in the U.S., as the software is updated to better identify hazardous water levels.
San Antonio served as the site of the specific failure that triggered the wide-scale recall [2]. The company is now working to refine how its AI interprets visual data when roads are submerged, a challenge that has plagued various levels of autonomous driving technology.
While the vehicle involved in the San Antonio incident was unoccupied [3], the potential for passenger injury in a similar scenario prompted the immediate recall of thousands of units. The company is reviewing the data from the event to implement safeguards that would force a vehicle to stop or reroute before encountering deep water.
“Waymo recalled approximately 3,800 robotaxis”
This incident underscores the 'edge case' problem in autonomous vehicle deployment, where rare but high-impact environmental events can bypass standard safety protocols. By recalling thousands of vehicles due to a single incident, Waymo acknowledges that its perception system lacks the necessary reliability to navigate flood-prone regions safely, potentially slowing the rollout of robotaxis in cities prone to extreme weather.
