Zoox Recalls 105 Robotaxis After Smoke Failure

Zoox has recalled software in 105 autonomous vehicles after an unoccupied robotaxi entered heavy smoke that concealed an active fire scene in Las Vegas.

No injuries were identified. The incident exposes a distinct challenge for driverless systems: emergency hazards are communicated through changing combinations of smoke, lights, cones, people and vehicle movement.

Smoke Obscured the Scene

On June 20, the Zoox vehicle encountered heavy smoke at an active emergency scene that had not yet been blocked with traffic cones, according to the company's report to federal regulators.

The vehicle entered the area, braked hard while attempting to steer away and stopped. A remote teleguidance employee then instructed it to reverse. First responders placed cones after the vehicle withdrew.

Zoox said its review found no other event of the same kind. The vehicle was empty, limiting direct passenger risk in this incident, but an obstruction could still delay emergency operations.

Software Update Followed

The voluntary recall covers 105 vehicles and addresses detection and response to heavy smoke. Because the defect is software-based, the remedy does not require replacing a conventional mechanical component.

An over-the-air correction can be deployed quickly, but the recall label remains important. It creates a public regulatory record and requires the company to explain the safety defect and remedy.

The incident also shows the value of remote support. Teleguidance helped the vehicle leave, but it acted after the automated system had already entered the smoke.

Regulators Widened the Focus

The National Highway Traffic Safety Administration issued a July 8 call to action after identifying multiple cases of driverless vehicles interfering with police, firefighters and ambulances.

The agency cited failures involving flashing lights, flares, smoke, fire and traffic cones. It told developers to prioritize first-responder interactions and said it would meet with driverless-system companies by the end of the month.

The statement did not single out one manufacturer. It framed emergency-scene behavior as an industry-wide competence that regulators expect every developer to address.

Emergency Cues Differ

Ordinary road driving often uses stable lane markings, mapped routes and predictable traffic controls. Emergency scenes are temporary and incomplete. Smoke may hide the hazard before cones arrive, while responders may redirect traffic by hand.

A safe system must combine perception with context. Detecting smoke as pixels is not enough if the vehicle does not infer that visibility loss may signal fire, responders or an obstructed road.

It must also choose a conservative action without stopping where it blocks access. That makes emergency response a planning problem as well as a sensor problem.

Scale Raises Stakes

Zoox offers free rides in parts of Las Vegas and San Francisco and has selected-user operations in small areas of Austin and Miami. Testing continues in other cities.

Each expansion increases the variety of temporary conditions the software may encounter. Rare events matter because fleet mileage turns low-probability situations into recurring operational tests.

Other autonomous-vehicle developers have also recalled software for unusual road interactions. The relevant measure is not whether recalls occur, but whether companies detect problems, disclose them, fix them and prevent repeat behavior.

TheTrendsWire’s Take

Sources

• NHTSA — Automated Vehicles and First Responders

• U.S. Transportation Department — Recall Data

Read More

• Waymo Teen Ride Raises Privacy Questions

• Hyundai and Kia Recall 14 EVs

• Subaru Recalls Vehicles Over Weight Labels

• Ford Recalls 420,000 Vehicles

• Lloyds Builds Autonomous AI Systems