← The Crash Report
Investigation

Tesla's Robotaxi Remote Operators Have a 100% Crash Rate in NHTSA Filings

Rex Driverton • Clautomotive Desk
September 4, 2026
A dark control room with a single monitor showing a Tesla Model Y stuck on a dead-end residential street at night

Tesla has filed 22 unique crash reports with NHTSA under its Standing General Order for automated driving systems since launching robotaxi service in Austin last June.[1] Three describe a remote operator taking direct control of the vehicle. All three ended in collisions.[2]

3 for 3
Every disclosed remote-operator takeover in Tesla's NHTSA robotaxi filings ended in a collision

July 2025: the autonomous system refuses to proceed, a teleoperator takes the wheel remotely, accelerates, turns left, drives up a curb into a metal fence at 8 mph. January 2026: navigational help requested, teleoperator takes over, drives into a temporary construction barricade at 9 mph. May 2026 in Houston: a Model Y stuck on a dead-end residential street, remote operator attempts recovery, hits a hidden tree stump at 2 mph.[2] That Houston filing is the first where Tesla coded its Driver/Operator Type as "Remote (Commercial / Test)" in structured NHTSA data. Earlier incidents only surfaced when Tesla unredacted all 17 crash narratives in May after a year of marking them confidential.[3]

Obvious caveat: crash reports exist because there was a crash. Tesla does not disclose how often remote operators intervene successfully, so a true failure rate is unknowable. But 3 for 3 is what federal filings show.

Waymo, operating roughly 3,000 robotaxis and completing over 500,000 paid trips per week, draws the line differently.[4] In a February 2026 letter to Senator Ed Markey, Waymo stated it "has not used remote driving or 'tele-operations' where a human performs the Dynamic Driving Task." Its remote agents provide guidance the vehicle can accept or reject. They do not steer. Waymo acknowledged one narrow exception: a U.S.-based event response agent can prompt a stuck vehicle forward at 2 mph with fixed steering angles, a function the company says "has never been used outside of training."[2]

Both companies built a tool for getting a stuck car unstuck. One has never deployed it on a public road. Tesla has deployed its version three times, and all three appear in crash reports filed with the federal government. Tesla's entire active fleet consists of 31 vehicles, only 14 operating without an in-car safety monitor, down from a peak of 25 unsupervised in late April.[5] Texas DMV records show 42 Tesla permits for driverless operation against 577 for Waymo.

If you ride in any robotaxi, human backup is part of every program. What differs is whether that human advises from a distance or steers over a cellular link with latency and no physical feel for the road. Check nhtsa.gov/technology-innovation/automated-vehicles-safety for the latest ADS incident data before assuming "driverless" means what you think it means.

Sources & References

  1. NHTSA, Standing General Order on Crash Reporting for ADS and Level 2 ADAS, SGO-2021-01. nhtsa.gov
  2. Fred Lambert, “Tesla remote operator crashed a ‘Robotaxi’ in Houston, NHTSA data shows,” Electrek, July 20, 2026. electrek.co
  3. Fred Lambert, “Tesla finally reveals what happened in 17 ‘Robotaxi’ crashes,” Electrek, May 15, 2026. electrek.co
  4. Fred Lambert, “Tesla ‘Robotaxis’ are not crashing because they are not running,” Electrek, June 16, 2026. Waymo fleet size from same analysis. electrek.co
  5. Texas Department of Motor Vehicles, Driverless Vehicle Registry, effective May 28, 2026; Robotaxi Tracker independent fleet data. electrek.co

Methodology note: Based on NHTSA Standing General Order filings for automated driving systems and publicly available fleet data. Remote operator crash record reflects only disclosed incidents; Tesla does not report total intervention counts. Waymo teleoperation comparison is from Waymo's February 2026 letter to Senator Ed Markey. All data is observational.