New data from the National Highway Traffic Safety Administration (NHTSA) is finally available to the public, and it paints a rocky picture for Tesla’s self-driving dreams. For months, Tesla fought to keep parts of this data hidden, telling regulators that the information contained “confidential business secrets.” However, unredacted reports now show that at least two Tesla robotaxi crashes happened while remote human operators were actually steering the vehicles.
These incidents occurred in Austin, Texas, where Tesla first launched its robotaxi service in June 2025. In both cases, a safety monitor sat in the driver’s seat, but no passengers were in the back. The first crash happened in July 2025. A safety monitor in the car got nervous and asked for help. A remote teleoperator took control from an office, but instead of stopping safely, the operator increased the vehicle’s speed. The car drove right up over a curb and smashed into a metal fence.
The second major incident happened more recently, in January 2026. Again, a remote operator assumed control of the vehicle. This time, the operator drove the robotaxi directly into a temporary barricade at a construction site. While the car was only moving at 9 MPH, the crash highlights a major flaw in the system. While other companies like Waymo use remote workers to give the car advice, Tesla’s system allows these workers to actually drive the car like a video game. This “remote driving” seems to be causing more problems than it solves.
The problems don’t stop with remote drivers. Tesla’s software has struggled on its own as well. Reports show at least two instances where robotaxis accidentally clipped the side mirrors of other parked cars. In a more heart-pounding moment, a robotaxi hit a dog that ran into the middle of the street. Luckily, the dog survived the encounter, but the car’s sensors failed to stop in time to avoid the impact. These “minor” mistakes suggest that the software still has trouble navigating the messy reality of city streets.
Even when the cars stay on the road, the service itself seems to be a mess for customers. A reporter from Reuters recently tried to use the service in Dallas on a Monday afternoon and had a terrible experience. The reporter wanted to go from the Southern Methodist University campus to Dallas City Hall. This is a 5-mile trip that usually takes about 20 minutes on a major freeway. Instead, the reporter waited nearly 2 hours to complete the short journey.
If waiting for two hours isn’t bad enough, the cars often fail to drop people off where they actually want to go. Passengers have reported multiple times that the robotaxi would pull over and end the ride 15 minutes away from their actual destination. Even though the desired drop-off point was clearly inside Tesla’s approved coverage area, the car simply refused to go the final few blocks. This forces passengers to walk the rest of the way, defeating the purpose of a taxi service.
Tesla has likely poured well over $1 billion into this technology over the last few years. The company desperately wants to catch up to Waymo, which currently leads the market. However, these reports show that Tesla is far behind. If a car cannot navigate a simple construction zone or drop a passenger at the correct door, it is not ready for mass use. Even a 1.5% failure rate is too high when you are dealing with multi-ton vehicles moving through crowded neighborhoods.
The federal government is keeping a very close watch on these developments. NHTSA requires all self-driving companies to report crashes, but Tesla’s attempt to redact the data has frustrated many safety advocates. They argue that the public has a right to know if the “driverless” car next to them is actually being controlled by a person in a remote office who might not have a clear view of the road.
Tesla’s rivals are also dealing with their own software bugs and occasional bumps, but they seem to have a smoother operation. Waymo’s cars, for example, have been driving in Phoenix and San Francisco for much longer without the same level of reported frustration over wait times and drop-off errors. For Elon Musk, the pressure is mounting to prove that his vision for an autonomous future isn’t just an expensive experiment.
For now, residents in Austin and Dallas are the ones testing this technology in their daily lives. They have to deal with the 2-hour wait times and the cars that hit fences. Tesla clearly has a very long way to go before its robotaxis become a reliable way to get around. Until the software can drive better than a remote operator and drop people off at their actual destinations, the service will likely remain a curiosity rather than a true competitor in the ride-sharing world.
