Waymo’s Robotaxis Are Getting Safer. Scaling Them Is Exposing New Problems

Waymo's robotaxis are becoming increasingly safe, but scaling their operations is revealing new challenges. Waymo's safety data indicates that their robotaxis are reducing crashes, yet recalls and emergency-scene failures suggest a more significant challenge when autonomous fleets encounter temporary road layouts, emergency scenes, and other unusual conditions.

Waymo recalled 3,871 autonomous vehicles on June 17 after discovering instances where their driving system entered freeway construction zones, highlighting this class of issues that becomes more critical as driverless fleets face unexpected road conditions. Despite outperforming human drivers in crash statistics, autonomous vehicles still struggle with rare situations that humans can typically interpret instinctively.

Waymo's safety data reveals that their vehicles have now driven 220.6 million rider-only miles without a human driver, resulting in 94% fewer crashes causing serious or fatal injuries compared to human-driven benchmarks. However, these figures do not account for the situation where an autonomous system may not handle every unusual scenario effectively.

In July, the National Highway Traffic Safety Administration (NHTSA) documented a pattern of driverless AVs interfering with law enforcement and first responders, such as entering emergency scenes, blocking ambulances, and failing to respond appropriately to emergency signals. Other companies like Zoox and Tesla have also faced similar problems with software recalls related to emergency situations.

As more robotaxis hit the road, the number of real-world experiences grows, but so does the potential for encountering something the autonomous system has not been trained to handle. Scaling autonomous vehicle deployment does not stop the need for real-world encounters to turn into reliable software improvements. The next phase of the robotaxi race will be judged not only by fleet size, miles driven, or the number of cities served, but by whether companies can turn real-world encounters into less frequent, severe, and repeatable issues.

If successful, autonomous driving can transform from an impressive demonstration into dependable transportation infrastructure.