Tesla's Full Self-Driving (FSD) system is not yet ready to operate safely without human intervention, according to large-scale real-world testing in Australia. Researchers conducted tests over 100 days on Queensland roads and documented more than 500 critical situations where the system either made dangerous errors or behaved unpredictably. The findings suggest that achieving truly autonomous vehicles will require several more years of development and millions of additional testing miles.

What Did the Australian Testing Reveal About Tesla FSD?

The research team created a special open archive called White Box Autonomy to record every failure and erratic behavior from a Tesla Model Y during real-world driving. While the system sometimes performs more precisely than professional drivers, it makes unexpected errors in situations that should be straightforward. The most alarming failures occurred in specific scenarios that highlight fundamental gaps in the system's understanding of road environments.

One of the most dangerous findings involved railway crossings. In one instance, the Tesla followed a car in front and nearly stopped directly on the tracks. The driver had to engage the emergency braking system to prevent a potential collision. This type of failure demonstrates that autonomous systems still cannot fully analyze their surroundings and anticipate hazards the way human drivers do.

Speed-restricted zones near schools proved particularly problematic. The system was wrong in 90 percent of cases when navigating these areas. Even more concerning, the car would unreasonably slow down in school zones even when school was not in session and restrictions were not in effect, hindering other drivers on the road.

Where Does Tesla FSD Struggle Most?

The testing revealed specific scenarios where the FSD system consistently fails to perform safely or intelligently. These challenges go beyond simple technical glitches; they reflect deeper limitations in how the AI interprets complex driving situations.

• Bridge Navigation: The system misinterprets well-marked road lines when crossing small bridges, causing the vehicle to "snake" across lanes unpredictably.
• Complex Intersections: The system struggles with roundabouts and has difficulty navigating intersections that require informal communication with other drivers, such as yielding or merging.
• Parking Situations: The vehicle cannot safely pass between cars parked densely on the side of the road, a common driving scenario in urban areas.
• Object Misclassification: The system misidentifies scooter riders as pedestrians, leading to incorrect driving responses.
• Weather Conditions: The FSD cannot reliably see road signs in poor weather, limiting its ability to respond to changing road conditions.

Researchers emphasized that the most difficult task for autonomous systems is engaging in informal communication with other drivers. Algorithms often hesitate in situations that require judgment calls about who has the right of way or how to safely merge into traffic. These are decisions that experienced human drivers make intuitively, but they remain challenging for current AI systems.

How to Understand the Gap Between Current FSD and Full Autonomy

• Auxiliary Function vs. Full Replacement: Autonomous driving systems are useful as helper features that assist drivers, but they are not yet safe for completely unsupervised operation without a human ready to take control.
• Data Requirements: Companies continue to improve their software through continuous testing and updates, but achieving full autonomy will require several more years of development and millions of additional kilometers of real-world driving data.
• The "ChatGPT Moment" Timeline: NVIDIA CEO Jensen Huang predicted a breakthrough moment for self-driving cars similar to the rapid advancement seen in large language models, but this milestone remains some distance away based on current testing results.

The Australian research directly contradicts the optimistic timelines that some industry leaders have suggested. While Tesla and other autonomous vehicle companies continue to make incremental improvements, the gap between current capabilities and safe, fully autonomous operation remains substantial.

What Is Tesla Doing to Address These Limitations?

Despite the testing failures documented in Australia, Tesla is advancing its robotaxi ambitions through the Cybercab, a purpose-built autonomous vehicle unveiled in 2024. The Cybercab is a two-seater vehicle that lacks a steering wheel or pedals and is intended for unsupervised full self-driving operation. Production is set to ramp up in 2026, with the company positioning it as a key part of a future ride-hailing network.

Notably, Tesla is also focusing on making its autonomous vehicles accessible to people with disabilities. At the National Federation of the Blind's annual convention in Austin, Tesla demonstrated Cybercab features designed specifically for blind and visually impaired passengers. The vehicle includes Braille lettering on physical controls, dedicated space for service animals and assistive devices, and seating at wheelchair height for easier transfers.

This accessibility focus addresses a significant gap in transportation options. In the United States, approximately 2.2 million people are blind or have significant vision impairment that affects daily mobility. For many, reliable, independent transportation remains a major barrier to employment, social engagement, and daily life. Tesla's emphasis on inclusive design could help provide mobility freedom to people who currently depend on paratransit services, which frequently face delays and availability issues.

The Cybercab builds on Tesla's extensive real-world data from millions of miles driven under its FSD program. The vehicle features advanced camera systems, neural network processing, and over-the-air updates designed to improve performance continuously. However, regulatory hurdles remain for widespread unsupervised robotaxi deployment, and Tesla continues to work with authorities in various states to address safety and operational concerns.

What Does This Mean for the Future of Self-Driving Cars?

The Australian testing results suggest that the timeline for fully autonomous vehicles may be longer than some industry optimists have claimed. While companies like Tesla, Waymo, and others continue to invest heavily in autonomous technology, the real-world complexity of driving remains a significant challenge. The gap between performing well in controlled conditions and operating safely in the unpredictable real world is wider than many expected.

For consumers and businesses waiting for robotaxi services, the message is clear: autonomous driving systems are improving, but they are not yet ready to operate completely without human oversight. The technology will continue to advance, but achieving the level of safety and reliability required for widespread deployment will take additional time, testing, and refinement. In the meantime, FSD remains most useful as an advanced driver assistance feature rather than a replacement for human drivers.