Austin Robotaxi Safety

Why Austin Matters

Austin, Texas is the epicenter of Tesla's autonomous driving ambitions. It is the only city where Tesla operates fully unsupervised Level 4 autonomous vehicles, meaning there is no human safety driver in the vehicle ready to take control. This makes Austin the proving ground for Tesla's vision of a scalable, driverless ride-hailing service.

While Tesla also operates robotaxis in the San Francisco Bay Area, those vehicles operate under California's supervised testing requirements with a safety driver behind the wheel. The Bay Area fleet serves as a data-collection and public demonstration operation, but it does not represent true autonomous performance.

Every safety metric on our main tracker is derived from Austin data specifically because it is the only deployment where the vehicle's autonomous system bears full responsibility for driving decisions.

Fleet Growth: From 10 to 80 Vehicles

Tesla's Austin fleet has grown significantly since launch:

This 8x growth in fleet size over approximately 7 months demonstrates Tesla's commitment to scaling the Austin operation. Importantly, safety has improved even as the fleet has grown, suggesting that fleet expansion does not degrade performance.

The fleet currently consists of modified Model 3 and Model Y vehicles. Tesla's purpose-built Cybercab, which will not have a steering wheel or pedals, is expected to begin production in April 2026 and may join the Austin fleet later that year.

Safety Monitor Removal: Current Status

The removal of in-vehicle safety monitors is one of the most closely watched milestones in Tesla's autonomous driving program. Here is the current situation:

• Some Austin vehicles already operate without safety monitors. Tesla began removing monitors from select vehicles in late 2025, starting with routes and conditions where the autonomous system had demonstrated the highest reliability.
• Most vehicles still have monitors present. The exact split between monitored and unmonitored vehicles is not publicly disclosed by Tesla.
• Monitors are observers, not drivers. Even when present, safety monitors in Austin sit in the passenger seat and do not have access to driving controls. They can request the vehicle pull over via a tablet interface but cannot directly steer or brake.

The presence of safety monitors creates a data interpretation challenge. If monitors occasionally intervene to prevent crashes (even by requesting a pullover), the observed crash rate reflects "autonomous system plus human backup" rather than the autonomous system alone. As more monitors are removed, the data will more accurately reflect pure autonomous performance.

What Conditions Must Be Met for Fleet-Wide Removal?

Tesla has not published explicit criteria for fleet-wide safety monitor removal, but based on regulatory requirements, industry precedent, and Tesla's public statements, the following conditions are likely necessary:

Sustained Safety Performance

The fleet must demonstrate MPI at or above human-driver benchmarks over an extended period. The most commonly cited target is the insurance-claim baseline of approximately 300,000 MPI. At the current improvement rate of doubling every ~79 days, this benchmark could be reached within a few months if the trend holds.

Statistical Significance

With a fleet of ~80 vehicles driving 115 miles per day each, the fleet accumulates roughly 9,200 miles per day. Achieving statistical confidence in safety metrics requires sustained operation over many weeks without significant incidents. A single bad week can shift the MPI considerably at current fleet sizes.

Regulatory Confidence

Texas does not currently require state permits for autonomous vehicle testing or deployment, making Austin a regulatory-friendly environment. However, Tesla likely coordinates with NHTSA and Texas DOT to maintain good standing. Any significant safety event could slow the removal timeline.

Remote Operations Capability

Tesla maintains a remote operations center that can communicate with vehicles and provide route guidance. Demonstrating that remote operators can effectively support unmonitored vehicles in edge cases (construction zones, unusual road conditions, system faults) is likely a prerequisite for full monitor removal.

Insurance and Liability Framework

Operating unmonitored vehicles requires appropriate insurance coverage and clear liability frameworks. Tesla must ensure that incidents involving unmonitored vehicles are covered and that responsibility is clearly assigned.

How Safety Is Tracked via NHTSA SGO

All incidents involving Tesla's Austin robotaxis are reported to NHTSA under Standing General Order 2021-01. This federal mandate requires reporting of any crash where the autonomous driving system was engaged within 30 seconds of the event.

Our tracker uses these SGO reports, combined with daily fleet size data from robotaxitracker.com, to calculate the miles per incident metric. The current trajectory shows exponential improvement, with MPI doubling approximately every 79 days.

Key safety observations for Austin as of early 2026:

• Zero fatalities since the program began
• Zero serious injuries across all reported incidents
• Most incidents are low-severity property damage events
• Exponential improvement trend with high R-squared fit

For complete details on methodology and data interpretation, see our FAQ page.

Current Safety Trajectory and What It Means

The data from Austin tells a compelling story of rapid improvement. Starting from single-digit thousands of miles per incident in the summer of 2025, the fleet has improved to approximately 92,500 MPI as of December 2025. This exponential trajectory, if sustained, has significant implications:

• Near-term (Q1 2026): MPI could reach 150,000-200,000 miles, approaching the lower bound of estimated human "all crashes" rates
• Mid-term (Q2-Q3 2026): MPI could reach 300,000-500,000 miles, matching or exceeding insurance-claim and police-report human baselines
• Longer-term: If exponential improvement continues, MPI could approach or exceed Waymo-level performance (~1M+ miles)

These projections assume the current improvement rate is maintained, which is not guaranteed. The trend could slow, plateau, or even reverse if new challenges emerge. That is why continuous tracking matters.

Austin's success or failure will determine the timeline for Tesla's expansion to other cities. If the safety trajectory holds through monitor removal and fleet scaling, expect accelerated launches in Dallas, Houston, Phoenix, and other announced cities.