Five Tesla Robotaxi crashes now share one detail: a remote operator, not the car’s own driving software, was in control when a Model Y met a fence, a barricade, a stump or a dumpster. Electrek editor Fred Lambert reported on Oct. 15 that every remote takeover appearing in a Tesla crash report so far has ended in a collision, bringing the total to five. Tesla has 36 unique Robotaxi incident reports on file with federal regulators, up from 22 in July.
Save more, waste less: The Cool Down’s newsletter sends smart, simple changes that help you and the planet in your everyday life.
We may earn a commission if you sign up.
None of the five involved an injury, and the speeds were low, from roughly 0.4 mph to 9 mph. The pattern still matters because the takeovers happened in the exact situations remote help exists to solve: a car that is stuck, boxed in or unsure how to proceed.
The two August filings from Dallas and Austin
The newest entries surfaced in the September release of automated-driving crash data under the Standing General Order. In report 13781-16678, filed from Dallas, a Model Y sat stopped on a residential street between a concrete mixer truck and a parked SUV. Tesla’s narrative says “The remote support operator remotely took control,” after which the car moved forward and touched the SUV at about 0.40 mph. Nobody was aboard, and the damage was to property only.
The second, report 13781-16679 from Austin, describes a remote operator driving a Model Y straight down an alley with a utility pole on one side and a dumpster on the other. The car struck the dumpster at about 1 mph. Tesla’s narrative says the operator was in control, and again there were no passengers and only property damage. Both events are dated August 2026, though the federal dataset publishes only month and year.
Three earlier crashes that set the pattern
Before those two, Electrek had counted three. Tesla unredacted its crash narratives in May, and the text showed what had been hidden. In July 2025 in Austin, a safety monitor could not get a car to proceed, so a teleoperator took over, accelerated, turned left and drove up a curb into a metal fence at 8 mph. In January 2026, also in Austin, a teleoperator stepped in for navigation help and drove into a construction barricade at 9 mph.
The third came from Houston, and Electrek’s July report on it carries the detail that matters most. In report 13781-15395, a Model Y stopped on a dead-end residential road was being guided out by a remote assistance operator when it rolled onto a grassy slope and hit a hidden tree stump at 2 mph. It was the first of Tesla’s 22 reports at that point to code the driver type as “Remote (Commercial / Test).” Three of the five crashes now carry that code.
What a crash-only dataset can and cannot show
Electrek’s claim deserves a careful reading, because the data is built from crashes. The federal Standing General Order requires named manufacturers and operators to report certain crashes within 30 seconds of an automated driving system being in use, with the most severe due within five days and less severe ones filed monthly. A remote takeover that ends safely never produces a report. So “every takeover in a crash report ended in a collision” is true almost by construction, and it says nothing about how many takeovers ended without one.
The missing number is how often Tesla’s operators step in at all. Tesla does not disclose it, so Lambert notes that no crash rate can be calculated, and NHTSA itself warns that the figures are not normalized for miles traveled. The agency also withholds the last six digits of each VIN, shows dates only by month and year, and cautions that initial reports can be incomplete or split one crash across several filings.
What the filings do establish is narrower and still worth knowing: when Tesla’s own narratives say a human was driving remotely, the vehicle was moving at walking pace or slower, in tight quarters, and made contact with something.
Waymo’s different line on remote driving
The contrast with its main rival is sharp. In a February 2026 letter to Senator Ed Markey, Waymo said it “has not used remote driving or ‘tele-operations'” in which a human performs the driving task, according to Electrek’s account. Waymo describes its remote staff as offering guidance that the onboard system can accept or reject, and says a narrow tool that nudges a stuck vehicle forward at 2 mph has never been used outside training.
That distinction explains why the Tesla reports draw attention. A remote operator who suggests a path leaves the software responsible for the car. An operator who steers puts a person with a camera feed and a network connection in charge of a vehicle in an alley or a cul-de-sac, which is a different safety case. One Electrek reader, Timothy Miller, argued in the comments that latency and thin sensor data make that arrangement riskier than the alternatives, a view the five reports neither prove nor refute.
Tesla’s report count is also climbing quickly. The jump from 22 to 36 unique incident reports between July and October covers every kind of Robotaxi event, not only remote ones. With the next monthly data release due, the thing to watch is whether the remote-operator tally stays at five or grows, and whether Tesla ever publishes how many takeovers occur per thousand miles.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
Save more, waste less: Subscribe to The Cool Down →
We may earn a commission if you sign up.
More from Morning Overview
