A robotaxi gets stuck, a human takes over remotely, and the car meets a fence. It is an awkward scene for anyone selling the future. Somewhere, a cat is watching the entire operation from a windowsill and wondering why humans made parking this complicated.
Newly public Tesla crash accounts describe two collisions during remote control. That is an interesting finding. Declaring the humans guilty and the robot innocent would require considerably more evidence.
What the two reports describe
Electrek’s May 15 examination of the disclosures covered 17 Tesla automated-driving incidents from July 2025 through March 2026. Its account identifies two Austin crashes involving remote takeover.
In July, a safety monitor requested assistance after the automated system stopped and would not proceed. According to Tesla’s narrative, the remote operator took control, turned left, went over a curb and struck a metal fence. The reported speed was 8 mph; the incident included a minor injury without hospitalization. No passengers were aboard.
In January, a monitor requested navigational assistance. The vehicle was stopped when the teleoperator took over, then travelled into a temporary construction barricade at 9 mph, damaging the front-left fender and tire. That report listed property damage.
The distinction matters: describing both events as injury-free would erase something the reporting actually records. Low speed also does not make a collision a successful recovery.
Remote control is part of Tesla’s recovery plan
Tesla explained its approach in a March 26 response to Senator Ed Markey. The company says vehicles are not normally driven remotely, but operators may temporarily assume direct control as a last resort after other interventions have been exhausted.
The letter describes takeover at speeds of 2 mph or less and a maximum speed authority of 10 mph once direct access is granted. It also describes advice, short repositioning commands and destination changes. Tesla says its operators work from Austin and Palo Alto.
Those are the company’s descriptions of its system and safeguards. They help explain how a remote takeover can happen; they do not independently establish what went wrong during either collision.
The useful mental picture is a service with a recovery procedure. A vehicle encountering a problem can call for help. The quality of that help belongs in the assessment of the service, just as the quality of the vehicle’s ordinary driving does. The recovery team does not become a separate universe the moment something gets scraped.
Waymo’s remote assistance means something different
Waymo’s February explanation says its remote staff respond to requests from the automated driver with information or advice. The system may accept or reject that advice and retains the driving task. Waymo says this role does not involve remote driving or continuous monitoring of every vehicle.
That is a meaningful difference from taking direct control. Lumping every human support function under “someone secretly drives the robot” makes the story simpler while making the explanation worse.
The recent Waymo traffic in an Atlanta cul-de-sac raised its own questions about how an automated fleet behaves around a neighbourhood. Different incidents can be worth discussing without assuming that every company uses the same machinery behind the curtain.
A crash report is a starting point, not a verdict
NHTSA cautions that these manufacturer-submitted reports can be incomplete or unverified. New information can lead to updates. An account identifying who had control at impact is valuable, but it does not settle every question about the interface, instructions, vehicle behaviour or circumstances leading to the handover.
The agency also warns against treating raw counts as comparable crash rates: the reporting does not normalize them for mileage, fleet size or operating conditions. Two remote-control incidents among a set of reports cannot tell us how often remote assistance is requested or how often it succeeds.
There are sensible questions to ask next. What could the operator see? What actions did the system permit? What review followed the incident? The public descriptions do not justify filling those gaps with an imagined bad connection, an inattentive employee or a flawless robot.
The backup deserves attention too
The strange part of this story is how ordinary its practical lesson feels. A sophisticated system still needs a way out when it gets stuck, and that way out needs to work.
Publishing more detail gives everyone something better to discuss than a scoreboard of crashes and brand loyalties. It invites scrutiny of the whole trip, including the moment the automated driver asks for assistance.
The cat remains unimpressed. Its own navigation system occasionally gets trapped behind an open door, but at least it has never described the rescue operation as a transportation revolution.
🐾 Curiosity looks good on you. Explore the Goodies, or find our illustrated books on Amazon.
