Zoox Recalls Self-Driving Cars After Smoke Detection Safety Issue
Amazon-owned autonomous vehicle company Zoox has announced a recall of its 105 self-driving vehicles after identifying a software issue that could prevent the vehicles from properly detecting heavy smoke, creating potential safety risks during emergency situations.
The recall follows an incident in June in which an unoccupied Zoox robotaxi encountered dense smoke surrounding an active fire scene. According to the company, the vehicle entered the affected area before braking abruptly while attempting to avoid the hazard. Under remote operator guidance, the vehicle safely reversed, allowing emergency responders to secure the scene with traffic cones.
Zoox said it has developed a software update designed to improve the vehicle’s ability to recognize and respond to heavy smoke, enhancing its existing perception system without requiring hardware changes.
The announcement comes as U.S. regulators increase scrutiny of autonomous vehicle safety. The National Highway Traffic Safety Administration (NHTSA) recently warned autonomous driving companies about a growing pattern of robotaxis interfering with emergency responders, including firefighters, ambulances, and police officers.
Regulators have documented multiple incidents involving autonomous vehicles that failed to correctly interpret emergency conditions such as flashing lights, road flares, smoke, fire, and temporary traffic control devices. According to the agency, these shortcomings present significant public safety concerns and require immediate attention from developers.
The issue extends beyond Zoox. Other autonomous vehicle operators, including Waymo, have also faced investigations following reports of robotaxis blocking emergency vehicles, entering active incident scenes, or violating traffic rules around stopped school buses.
These incidents highlight one of the remaining challenges in autonomous driving technology. While self-driving systems have made significant progress in handling everyday traffic, rare and unpredictable scenarios—often referred to as “edge cases”—remain difficult for AI-powered perception systems to interpret correctly.
As autonomous vehicles continue expanding into public roads, regulators are placing greater emphasis on how these systems interact with emergency personnel. Successfully recognizing dynamic emergency environments is increasingly viewed as a critical requirement before large-scale commercial deployment can become commonplace.



