Waymo Recalls 3,791 Robotaxis Over Flooded Road Detection Failure

Waymo, the self-driving car company owned by Alphabet, recalled 3,791 autonomous vehicles after one of them drove onto a flooded road during heavy rains in San Antonio. The unoccupied vehicle entered a section of roadway with standing water despite software systems that should have detected and avoided the hazard. Waymo reported the problem to the National Highway Traffic Safety Administration (NHTSA) on May 1, and the agency formally acknowledged it on May 11.

The incident happened when severe weather caused flooding in San Antonio. A Waymo vehicle failed to detect the flooded road and drove into it at speeds up to 40 mph. The exact nature of the software flaw has not been publicly detailed, but it clearly broke the systems meant to keep the car away from water-covered roads.

How Waymo Responded

Waymo deployed a software update to all 3,791 affected vehicles across its entire operating fleet. These cars serve paying customers in Phoenix, San Francisco, and Los Angeles, plus test vehicles in other cities. The fix was delivered remotely over-the-air—the same way your phone gets updates—rather than requiring vehicles to return to a service center.

This type of recall is routine in the autonomous vehicle industry. Companies can fix many safety problems with software patches without physical repairs. The ten-day gap between Waymo's report and NHTSA's acknowledgment follows standard federal review timelines for autonomous vehicle safety updates.

Why Flooded Roads Are a Hard Problem

For self-driving cars, detecting flooded roads is surprisingly tricky. These vehicles rely on cameras and lidar—a laser-based sensing system that builds a 3D picture of the road ahead—to understand their surroundings. The challenge: standing water can fool these sensors in ways that are hard to predict. A shallow puddle and a deep flooding zone might look similar to a camera or lidar, especially as water levels rise and fall during heavy rain.

Traditional rule-based safety systems (essentially if-then decision trees) struggle with the dynamic nature of flooding. Water conditions change minute by minute, and road surface visibility can be compromised. The system needs to distinguish between hazardous conditions and normal driving situations—a problem that mirrors human drivers making the same judgment call, but one that algorithms must solve consistently and without error.

A Pattern Across the Industry

This is not Waymo's first edge case—a situation outside normal operating conditions. Earlier this year, Waymo's vehicles blocked streets in San Francisco during a major power outage because they couldn't operate under those unusual circumstances. The flooding incident reveals a similar vulnerability: real-world conditions that lab testing and controlled pilots didn't fully account for.

Across the autonomous vehicle industry, weather and environmental edge cases remain a persistent challenge. Snow obscures road markings, heavy rain degrades sensor performance, extreme heat stresses electronics. Each of these conditions tests the limits of current self-driving systems. Flooding represents one particularly difficult case because it carries both perception challenges (detecting the water) and safety risks (not knowing water depth or road integrity underneath).

The broader context here is that initial deployments of autonomous vehicles tend to work smoothly in controlled conditions and good weather, but real-world complexity surfaces vulnerabilities that need fixing. We have seen this pattern repeatedly across technology adoption: early systems refine through operational learning. Each problem discovered and solved makes the technology more robust overall.

What This Means for Safety and Expansion

The incident involved an unoccupied vehicle, which raises a practical question: How do autonomous vehicle operators decide when weather is too severe to operate. Most companies do suspend service during extreme conditions, but the thresholds and timing vary. Waymo's interim software update—likely more cautious flood-avoidance parameters—buys immediate safety while the company develops a more sophisticated permanent fix.

Fleet-wide recalls like this one show the scale of Waymo's ambitions. Managing thousands of vehicles across multiple cities, each with different weather patterns and road conditions, introduces operational complexity. As these fleets grow, robust safety validation and incident response processes become increasingly critical.

The transparency Waymo has shown—voluntary disclosure and public recall—matters as the autonomous vehicle industry faces closer regulatory attention. Competitors like Cruise have also faced safety-related suspensions. In this environment, maintaining trust with regulators and the public is essential for sustained growth.

Looking ahead, this incident will likely influence how federal regulators develop safety standards for autonomous vehicles, particularly around environmental hazard detection. The regulatory process demonstrated here—where companies report issues and NHTSA oversees fixes—will probably become the template for handling similar problems as more self-driving cars enter service. Individual incidents like this one are operational setbacks, but they also drive systematic improvements that move the technology closer to handling all conditions reliably.