The Human Bottleneck: Analyzing Tesla's Robotaxi Teleoperation Failures
The path to fully autonomous ride-hailing is rarely a straight line, often punctuated by "edge cases" that challenge the limits of both artificial intelligence and human intervention. For Tesla, the ambition of a scalable Robotaxi network has recently met a tangible hurdle: the human teleoperator.
Newly unredacted crash reports submitted to the National Highway Traffic Safety Administration (NHTSA) have revealed that Tesla Robotaxis have crashed at least twice while being remotely piloted. These incidents, occurring in Austin, Texas, shed light on the operational frictions Tesla faces as it attempts to transition from a driver-assist system to a fully autonomous fleet.
The Teleoperation Fail-Safe: A New Point of Failure
Tesla has previously informed lawmakers that it employs remote operators to pilot vehicles in specific, high-risk scenarios—specifically when a vehicle is in a "compromising position" and needs to be moved to avoid blocking traffic or awaiting a first responder. To maintain safety, Tesla limits these remote operations to speeds under 10 miles per hour.
However, the NHTSA data reveals that this safety mechanism can itself become a source of error. Two specific incidents highlight this vulnerability:
- July 2025: After the Automated Driving System (ADS) struggled to move forward while stopped, a safety monitor requested remote assistance. The teleoperator took control to move the vehicle toward the left side of the street but inadvertently drove the car up a curb and into a metal fence.
- January 2026: In another instance, a safety monitor requested navigation support. The teleoperator took control of a stopped vehicle and proceeded straight, subsequently colliding with a construction site barricade at approximately 9 MPH, damaging the front-left fender and tire.
In both cases, a safety monitor was present behind the wheel, and no passengers were onboard, preventing potential injuries.
A Pattern of Low-Speed Collisions
Beyond teleoperation errors, the unredacted reports provide a narrative for 17 total crashes recorded by Tesla's nascent Robotaxi network since last year. While many of these involved other vehicles crashing into the Tesla Robotaxis—a common trend seen in other AV companies like Waymo—several incidents point to systemic software struggles:
- Environmental Obstacles: In September 2025, a Robotaxi failed to avoid a dog running into the street.
- Infrastructure Failures: Another September 2025 incident involved a vehicle making an unprotected left turn into a parking lot and colliding with a metal chain.
This latter issue mirrors a known industry-wide challenge. The NHTSA recently closed an investigation into Tesla's Full Self-Driving (FSD) software's tendency to hit parking lot bollards and chains, and Waymo similarly issued a recall last year for low-speed collisions with gates and chains.
The Scaling Gap
When compared to competitors, Tesla's current scale is significantly smaller. While Waymo is reportedly conducting hundreds of thousands of rides per week, Tesla's network remains in a cautious, limited rollout.
Elon Musk has acknowledged that safety is the primary bottleneck for expansion, stating that the company is being "very cautious" to ensure the system is completely safe. The unredacted NHTSA data suggests that this caution is warranted; the transition from "mostly autonomous" to "fully autonomous" requires solving not just the AI's perception of the world, but the reliability of the human-in-the-loop systems used to rescue the AI when it fails.
Technical and Ethical Considerations
The revelation of these crashes has sparked significant debate among technical observers regarding the infrastructure of teleoperation. Critical questions remain regarding the latency and location of these operators. As one observer noted on Hacker News, the reliability profile of the system changes drastically based on internet quality and the geographical location of the staff:
"It would be fascinating to know where the remote drivers were located... This can change the reliability profile quite and bit."
Furthermore, the reliance on remote operators raises fundamental questions about the safety standards of autonomous transit. Critics argue that if similar remote-control experiments were attempted in commercial aviation, the public outcry would be immense, yet the automotive industry is currently treating the public road as a laboratory for these hybrid human-AI control schemes.