Zoox recalled all 105 of its self-driving taxis on July 17, 2026, after one of them drove into thick smoke from an active fire and could not figure out how to get through it safely.

The incident happened on June 20, 2026. A Zoox taxi with no passengers inside drove into heavy smoke from an emergency fire. Firefighters had not yet blocked off the road with traffic cones. The taxi braked hard, tried to steer away, and stopped. A Zoox staff member working remotely then backed the taxi away from the scene so first responders could place cones. Zoox told the National Highway Traffic Safety Administration, or NHTSA, that no one was hurt. The recall report does not say where the incident happened. TechCrunch

Zoox investigated what went wrong and told NHTSA this was the only event of its kind the company has seen. Through late June and early July, Zoox had several conversations with NHTSA about how serious the problem was, how often it happened, and what caused it. The company decided on July 7 to issue the recall. The fix was a software update sent wirelessly to all 105 taxis.

The recall came one day before NHTSA's top official, Jonathan Morrison, sent a letter to all self-driving car companies on July 8. He told them to stop interfering with first responders. "The inability to detect and appropriately respond to such situations represents a functional insufficiency," he wrote. He also said: "Emergency scenes are not rare or extreme 'edge cases.'" TechCrunch

Self-driving car companies often call situations their cars struggle with "edge cases," meaning unusual events that are hard to predict. Morrison's point is that fire scenes, emergency vehicles, and blocked roads are not unusual at all. They happen every day on public roads.

This is Zoox's fourth recall. Earlier recalls dealt with hard braking in March 2025, a crash with a passenger car in May 2025, and a collision with an e-scooter, also in May 2025. All four recalls involved a software problem found after the cars were already on the road, and all four were fixed with wireless updates rather than taking the cars to a shop.

Zoox is currently giving free rides in Las Vegas and San Francisco ahead of a planned commercial launch. That launch needs NHTSA to grant a special exemption, because Zoox's taxis have no steering wheel and no pedals. The recall and the attention it brings from regulators could make that approval harder.

The incident shows both a strength and a weakness in how these taxis work. When the self-driving system could not handle the situation, a remote human operator stepped in and moved the car. No one was hurt. The vehicle stopped instead of driving into danger. But a taxi that stops in the middle of an emergency scene, blocking firefighters until a remote operator can fix things, is still a taxi that needs to get better. NHTSA says the goal is not just avoiding injury this time, but making sure the cars can detect and respond to emergency scenes on their own.

The reason heavy smoke is so hard for a self-driving car is that it interferes with all the sensors the car uses to see the road. Cameras lose the visual detail they need. Lidar, which bounces laser light off objects to measure distance, gets scattered by smoke particles. Radar can see through smoke better, but may not provide enough information for the car to understand that it is looking at an emergency. So the car can end up in a situation where none of its sensors gives it enough information to make a good decision.

The bigger question NHTSA is asking the industry is whether self-driving systems can truly understand a whole scene, or whether the current approach of spotting individual objects and following a set of rules will keep failing in messy but common situations.