Tesla’s remote operators, the people who can take over a “Robotaxi” when its software gets stuck, have now appeared in several federal crash reports. Every one of those takeovers in the public data ended in contact with an object. The crashes were low-speed and no injuries were reported, but the pattern matters because remote assistance is the safety net that driverless services rely on.
What the reports say
Tesla reports crashes involving its automated driving system to the US safety regulator, NHTSA, under a standing general order. For a long time Tesla redacted the narrative text of those reports as confidential business information. According to Electrek, Tesla reversed that in May 2026 and unredacted all 17 crash narratives it had filed up to then. That release is what exposed the role of remote operators.
- July 2025, Austin: a teleoperator drove the vehicle up a curb and into a metal fence at about 8 mph.
- January 2026, Austin: a teleoperator drove into a construction barricade at about 9 mph.
- May 2026, Houston: a remote operator was driving a Model Y out of a dead-end residential road when it entered a grassy slope and hit a hidden tree stump at about 2 mph. Damage was property-only, with no airbag deployment and no tow.
- August 2026, Dallas and Austin: two more remote-operator crashes appeared in NHTSA’s September data release. In Dallas a stopped Model Y touched a parked SUV at about 0.4 mph after a remote operator took control; in Austin a Model Y under remote control hit a dumpster in an alley at about 1 mph. Neither had passengers.
Electrek’s October count is 36 unique Robotaxi incident reports in total, up from 22 in July, and five remote takeovers in the crash data, all of which ended in contact. Figures come from Electrek’s reading of NHTSA’s data; check the NHTSA standing general order page for the raw filings.
What this does and does not show
Tesla does not publish how often its remote operators step in, so nobody outside the company can turn these crashes into a rate. Five incidents could be a small share of thousands of interventions or a large one. The reports also say nothing about whether a human behind the wheel would have done better. What they do show is that remote driving, not only the autonomous software, is a source of crashes.
Every incident listed above was property-damage-only at low speed, and Electrek reports no injuries.
How Waymo describes its approach
In a February 2026 letter to Senator Ed Markey, as summarised by Electrek, Waymo said it does not use remote driving. Its remote staff provide guidance that the vehicle can accept or reject. Waymo’s one reported exception is a roughly 2 mph nudge at fixed steering angles, which Electrek says has reportedly never been used outside training. Tesla’s design, where an operator can directly drive the car, is a different approach. For more on how Waymo measures safety, see our Waymo safety benchmark piece, and for the commercial side of the robotaxi race see Go’s IPO analysis.
What to watch next
- Whether Tesla discloses how often operators take control, which would allow a real crash rate.
- Whether NHTSA’s next data releases add more remote-operator crashes.
- Whether regulators set rules for remote operators; none are confirmed in the sources used here.
Corrections to an earlier version
An earlier version of this article described crashes on May 12 in Austin and May 14 in Los Angeles, latency figures, a Cybercab prototype, insurance cost estimates and pending teleoperator licensing laws. None of that is supported by the reports cited here, and it has been removed.
Sources
- Electrek, Oct 15 2026: Tesla remote operators keep crashing its Robotaxis
- Electrek, Jul 20 2026: Tesla remote operator crashed a Robotaxi in Houston
- NHTSA: Standing General Order on crash reporting
Like our coverage? Add stepphase.com as a preferred source on Google to see more of our stories in Top Stories.
