Tesla Reveals New Details About Robotaxi Crashes—and the Humans Involved

For more than a year, Tesla has shielded details about its robotaxi crashes from public view. Now, the company has published new details in a federal database about 17 incidents, which took place between July 2025 and March 2026. In at least two of them, Tesla’s human employees appear to have played a hand in the crashes by remotely driving the otherwise autonomous cars into objects on the street.In both crashes, which happened in Austin, “safety monitors” were in the vehicles’ passenger seats to oversee the still-fledgling self-driving tech, and no passengers were riding in the cars. Both crashes occurred at speeds below 10 miles per hour. The new details were first reported by TechCrunch.In one incident, which took place in July 2025, the safety monitor experienced “minor” injuries after a remote worker drove the Tesla up a curb and into a metal fence at 8 mph. The monitor, who had requested help from Tesla’s remote driving team after the car stopped on the side of a street and wouldn’t move forward, was not hospitalized, Tesla reported.The other incident, in January 2026, happened after a safety monitor requested navigation help from the remote team. The remote driver took control and drove the car straight into a temporary construction barricade at 9 mph. The crash left the robotaxi's front left fender and tire scraped up, but Tesla didn’t report any injuries.Tesla, which does not have a public relations team, did not respond to WIRED’s request for comment.The new details draw attention to an often misunderstood but safety-critical part of autonomous vehicle operations: the human backstops who remotely monitor the robot cars and intervene when they get into trouble. All US self-driving operators maintain these remote teams, according to letters submitted to a US senator earlier this year. But Tesla appears to be an outlier because it more frequently allows these remote workers to directly drive the cars.Other companies typically allow their workers to remotely provide input to the autonomous vehicle software, which the system can choose to use or reject. (Waymo says that specially-trained workers can remotely drive its cars up to 2 mph, but said in February that it hadn’t used that functionality outside of training.)Safety advocates have raised questions about remote driving, which can be challenging in places without consistent cellular connectivity and in contexts where remote drivers need a perfect understanding of a car’s surroundings to guide it out of complex situations.The new details on the two Tesla crashes “raise questions about what the teleoperator can see in both coverage and resolution, and what kind of latency they are experiencing while driving,” Noah Goodall, an independent self-driving vehicle researcher, tells WIRED in a message.Tesla’s still-fledgling robotaxi service is operating in three Texas cities: Austin, Dallas, and Houston. But the service has fewer than 100 vehicles operating in total, compared to Waymo’s nearly 4,000. Less than half of Tesla’s cars appear to operate without a safety monitor sitting in the passenger seat. Reuters reported this week that service wait times in Houston and Dallas, where robotaxis launched in April, are upward of 35 minutes. Even in Austin, where the cars have been carrying passengers for almost a year, a reporter for the publication found that robotaxis were sometimes completely unavailable.Tesla CEO Elon Musk has said that autonomous vehicles and robotics are the automaker’s focus instead of manufacturing electric cars. Musk’s compensation—a potential $1 trillion paycheck by 2035—is now tied to vehicle and robot deliveries, as well as sales of not-yet-released self-driving subscriptions and the number of robotaxis in commercial operation.Autonomous vehicle operators are required by law to update the federal autonomous vehicle crash database, maintained since the Biden administration by the US Department of Transportation's National Highway Traffic Safety Administ