Tesla has disclosed a series of incidents involving its robotaxi fleet that reveal a recurring and critical failure point: the humans tasked with keeping the cars safe. New data submitted to a federal database details 17 incidents occurring between July 2025 and March 2026, highlighting a troubling trend where the “human backstop” intended to prevent accidents became the cause of them.
The records indicate that in at least two of these Tesla robotaxi crashes, remote employees took direct control of the vehicles and drove them into stationary objects. These incidents, both occurring in Austin, Texas, underscore the inherent risks of teleoperation—the process of driving a vehicle from a remote location—and raise questions about the resolution and latency of the systems Tesla uses to bridge the gap between a remote operator and the physical street.
In both instances, the vehicles were not carrying passengers, but they were equipped with safety monitors sitting in the passenger seats. These monitors are intended to oversee the autonomous software and intervene if the system falters. However, in these cases, the intervention came from a remote team, and it was that human intervention that led to the collisions.
The Austin Incidents: When Remote Help Fails
The first reported crash occurred in July 2025. After a robotaxi stopped on the side of a street and failed to move forward, the on-board safety monitor requested assistance from Tesla’s remote driving team. A remote worker took control of the vehicle but drove it up a curb and into a metal fence at approximately 8 mph. Tesla reported that the safety monitor sustained minor injuries but did not require hospitalization.
A similar failure occurred in January 2026. In this instance, a safety monitor requested navigation assistance. The remote driver took control of the car and drove it directly into a temporary construction barricade at 9 mph. While the collision scraped the front left fender and tire, no injuries were reported.
These events highlight a specific operational vulnerability. Unlike the autonomous software, which relies on a suite of sensors and real-time processing, a remote driver relies on a video feed. Any lag in the connection—known as latency—or a lack of depth perception in the camera feed can turn a simple navigation correction into a collision.
Direct Control vs. Software Input
The Tesla approach to remote intervention appears to be an outlier in the autonomous vehicle industry. Most operators of self-driving cars utilize remote teams not to “drive” the car in the traditional sense, but to provide high-level guidance that the car’s own software then executes.
For example, Waymo typically allows remote workers to provide input that the autonomous system can either accept or reject based on its own sensor data. While Waymo has noted that trained workers can remotely drive cars at speeds up to 2 mph, the company has stated that this functionality is largely reserved for training rather than active commercial operation. Tesla, conversely, allows its remote workers to exercise more direct control over the vehicle’s movements.
| Feature | Tesla Robotaxi Approach | Industry Standard (e.g., Waymo) |
|---|---|---|
| Remote Intervention | Direct steering and acceleration | Software-based guidance/input |
| Remote Speed | Capable of street speeds | Strictly limited (often < 2 mph) |
| Human Presence | Frequent use of safety monitors | Predominantly driverless |
Safety researchers argue that this direct-drive model is perilous. Noah Goodall, an independent researcher specializing in self-driving vehicles, noted that these crashes raise significant questions regarding what a teleoperator can actually see in terms of coverage and resolution, and the specific latency they experience while attempting to navigate a vehicle through a complex urban environment.
Operational Struggles in Texas
While Tesla has expanded its robotaxi service into Austin, Dallas, and Houston, the scale of the operation remains modest compared to its competitors. Tesla currently has fewer than 100 vehicles in operation across these three cities. In contrast, Waymo operates a fleet of nearly 4,000 vehicles.
The fledgling service is also struggling with availability, and reliability. In Dallas and Houston, where the service launched in April, wait times for a vehicle have been reported as exceeding 35 minutes. Even in Austin, where the service has been active for nearly a year, the cars are frequently unavailable to users.
Tesla’s reliance on human oversight remains high. Less than half of the vehicles in the Texas fleet currently operate without a safety monitor in the passenger seat, suggesting that the company is not yet confident in the software’s ability to handle the “edge cases” of Texas traffic without a physical human presence.
The High Stakes of Autonomy
The push toward full autonomy is not merely a technical goal for Tesla; it is a financial imperative for its leadership. CEO Elon Musk has explicitly shifted the company’s primary focus away from the mass manufacture of electric cars and toward robotics and autonomous transport.
This pivot is reflected in Musk’s compensation structure. A potential payout of up to $1 trillion by 2035 is tied directly to the delivery of robots and vehicles, the number of robotaxis in commercial operation, and the sales of self-driving subscriptions that have yet to be released to the general public.
As Tesla continues to refine its federal safety reporting and operational protocols, the tension between rapid deployment and rigorous safety remains. The transition from a monitored fleet to a truly driverless one requires solving not just the AI’s logic, but the failures of the humans who stand behind it.
The company has not provided a timeline for when it expects to remove safety monitors from the majority of its Texas fleet. The next significant checkpoint will be the upcoming quarterly safety filing, which is expected to provide updated incident rates for the spring 2026 period.
Do you think remote driving is a necessary safety net or a liability? Share your thoughts in the comments below.
