Tesla's robotaxi program has experienced at least two significant crashes since launching tests last year, according to newly unredacted government data, with both incidents directly caused by remote operator decisions that made dangerous situations worse. The National Highway Traffic Safety Administration (NHTSA) released crash documentation showing incidents in July 2025 and January 2026, both occurring in Austin, Texas.

What Actually Happened in Tesla's Robotaxi Crashes?

The first crash occurred in July 2025 when a safety monitor was present in the vehicle but required assistance. A remote operator took control of the robotaxi, but instead of stabilizing the vehicle, the operator increased its speed, causing it to drive up onto the curb and strike a metal fence. The second incident happened in January 2026, when a remote operator assumed command and directed the vehicle into a temporary construction site barricade while traveling at nine miles per hour. In both cases, safety monitors were behind the wheel and no passengers were aboard the vehicles.

What distinguishes these incidents from typical autonomous system failures is that the remote operators' actions directly caused or worsened the crashes. In the first case, increasing speed when a vehicle needed stabilization represents a decision that contradicted what the situation required. In the second case, the operator steered the vehicle into an obstacle rather than around it. These are not autonomous system failures; they are human decision-making failures happening in real time.

Why Do Remote Operators Sometimes Make Crashes Worse Instead of Better?

The crashes documented in Tesla's data suggest a fundamental challenge with using remote human operators as a safety mechanism. A remote operator sitting in a control center cannot see what the vehicle's sensors see, cannot feel the vehicle's motion, and must make critical decisions based on a video feed and telemetry data alone. When a vehicle transitions from autonomous to remote-operated control, the operator must instantly understand the vehicle's current state, trajectory, and the reason for the handoff.

In both Tesla incidents, operators appear to have made decisions that worsened the situation rather than stabilizing it. This may indicate that remote operation, while intended as a safety net, could introduce new failure modes that autonomous systems operating independently would not create. The gap between what an operator can perceive remotely and what the vehicle's sensors actually detect in real time may be creating a dangerous mismatch.

How Does This Compare to Other Robotaxi Services?

Waymo, one of the world's leading self-driving ride-hailing platforms, has also experienced its own challenges and accidents over the past few years. However, the source material does not provide detailed information about Waymo's remote operation architecture or how frequently Waymo relies on remote human intervention during normal driving scenarios. The source only notes that Waymo "has also seen its fair share of problems, glitches, and accidents," without specifying whether those incidents involved remote operator errors or autonomous system failures.

What the comparison does suggest is that different robotaxi services are taking different approaches to safety. Tesla's model appears to include remote operator intervention as a regular part of its safety strategy, while other services may rely more heavily on autonomous decision-making. The question of which approach is safer remains open, but Tesla's crash data indicates that remote operator intervention carries its own risks.

Steps to Understand Robotaxi Safety Differences

• Intervention Frequency: Determine how often a robotaxi service requires remote operator intervention during normal operations, as services with higher intervention rates may face more operator-error incidents like those documented in Tesla's crashes.
• Crash Pattern Documentation: Review whether documented incidents involve autonomous system failures or remote operator errors, since Tesla's crashes both involved remote operator actions that worsened the situation.
• Operator Training Standards: Consider whether operators have adequate training for split-second decisions and whether communication delays between the control center and vehicle could affect response times in critical moments.
• Passenger Safety Records: Examine whether services have experienced passenger injuries from remote operator errors, which is a critical factor as these services expand to more cities.

The slow progress of Tesla's robotaxi program, as indicated by the crash data and ongoing challenges, reflects the broader difficulty of scaling autonomous ride-hailing services. It is normal for self-driving services to encounter obstacles and problems as they develop their technology, but the nature of those problems matters significantly. When crashes result from remote operator decisions rather than autonomous system failures, it may suggest that the underlying technology is not yet ready for the responsibility being placed on it.

Companies developing self-driving technology must ensure their systems are in perfect condition to avoid compromising passenger safety. For Tesla, this means either improving the reliability of its autonomous decision-making so remote intervention is rarely needed, or fundamentally redesigning how remote operators are trained and equipped to make split-second decisions. The current model, as evidenced by these crashes, may be creating new safety problems rather than solving existing ones.

As robotaxi services expand to more cities, the architectural decisions made now about how to handle safety-critical moments will determine which companies succeed at scale. Tesla's crash data suggests that relying on remote operators to solve safety problems may introduce new risks that autonomous systems operating independently would not create.