Quick Verdict: Tesla FSD Is Not Self-Driving
If you're asking "is Tesla self-driving safe?" the first thing to understand is that Tesla's Full Self-Driving (Supervised) is not a self-driving system. It is an SAE Level 2 driver-assistance feature that requires you to remain fully engaged, hands near the wheel, and eyes on the road at all times. True self-driving robotaxis — like Waymo's Level 4 service — operate without any human driver in the vehicle, but they are geofenced to specific cities and conditions.
This distinction isn't semantic. It's the difference between a system that helps you drive and one that drives for you. The safety records, regulatory oversight, and real-world capabilities of these two categories are fundamentally different. Let's break down what the data actually shows.
SAE Automation Levels: Why the Distinction Matters
The Society of Automotive Engineers (SAE) defines six levels of driving automation. Understanding where Tesla FSD Supervised and Waymo sit on this scale is essential before evaluating safety.
| Level | Name | Who Drives? | Example Systems |
|---|---|---|---|
| 0 | No Automation | Human does everything | Standard vehicles |
| 1 | Driver Assistance | Human + one assist (steering OR speed) | Adaptive cruise control, lane-keep |
| 2 | Partial Automation | Human + combined assists (steering AND speed) | Tesla Autopilot, FSD Supervised, GM Super Cruise, Ford BlueCruise |
| 3 | Conditional Automation | System drives in limited conditions; human must take over when requested | Mercedes DRIVE PILOT (approved in select states, low-speed highway only) |
| 4 | High Automation | System drives without human in defined operational domain | Waymo One, Zoox (geofenced urban areas) |
| 5 | Full Automation | System drives everywhere, all conditions | Does not yet exist commercially |
Is Tesla FSD Supervised Safe? The Data
Tesla publishes quarterly safety reports claiming approximately one crash per 4-6 million miles driven with Autopilot engaged, compared to the U.S. average of roughly one crash per 670,000 miles (per NHTSA data). On the surface, that looks impressive. But there are critical caveats:
- Self-selection bias: Tesla's Autopilot data is predominantly highway miles in fair weather — the safest driving conditions. It does not represent a like-for-like comparison with the national average, which includes urban intersections, night driving, and adverse weather.
- Engagement reporting: Tesla's metrics track when Autopilot is active, but the system's engagement criteria (clear lane markings, divided highways) naturally filter for lower-risk scenarios.
- NHTSA Standing General Order: Under NHTSA's mandatory crash-reporting rule, Tesla has reported hundreds of incidents involving Autopilot and FSD. According to NHTSA data compiled through 2024, Tesla vehicles accounted for a significant share of reported Level 2 crashes — though this partly reflects Tesla's volume and reporting compliance rather than a uniquely high failure rate.
- FSD (Supervised) specific data: Tesla has not published granular, independently verified crash rates specifically for FSD in city-street mode separate from highway Autopilot. The system's behavior on complex urban routes — unprotected left turns, construction zones, irregular intersections — introduces edge cases that highway-only data doesn't capture.
- Tesla Autopilot claim: ~1 crash per 4-6 million miles (highway-biased, manufacturer-reported)
- U.S. average (NHTSA): ~1 crash per 670,000 miles (all conditions, all drivers)
- Waymo: ~7.1 million fully autonomous miles driven in California (2024 DMV data) with a lower police-reported crash rate than human benchmarks in comparable urban environments
- NHTSA investigation: Ongoing into Autopilot/FSD behavior near emergency vehicles and crash-site scenes
Waymo Robotaxi: How a True Level 4 Service Compares
Waymo One, Alphabet's robotaxi service, operates without any human behind the wheel. As of early 2026, Waymo has completed over 10 million paid, driverless trips across its service areas. Here's how it compares to Tesla FSD Supervised on safety, availability, and cost.
Where Can I Use Waymo Right Now?
| City | Status | Service Area Notes |
|---|---|---|
| Phoenix, AZ | Fully operational | Largest service area; covers most of metro Phoenix including airport runs |
| San Francisco, CA | Fully operational | City-wide coverage; 24/7 service |
| Los Angeles, CA | Fully operational | Expanded coverage zone across central LA and Westside |
| Austin, TX | Operational (expanding) | Initial zones live; geographic coverage growing |
Waymo's service is geofenced — the car will only operate within its mapped, approved zones. If you request a ride outside the boundary, the app won't let you book it. This is a deliberate safety constraint, not a limitation of the technology per se.
How Much Does Waymo Cost vs. Uber/Lyft?
| Service | 5-Mile Urban Trip | 10-Mile Urban Trip | Airport Run (15-20 mi) | Surge Pricing? |
|---|---|---|---|---|
| Waymo One | $10–$14 | $18–$26 | $35–$50 | Rare; more predictable pricing |
| UberX | $12–$18 | $20–$32 | $40–$65 | Yes, frequent in peak hours |
| Lyft Standard | $11–$17 | $19–$30 | $38–$60 | Yes |
Waymo pricing is generally competitive with UberX and Lyft, sometimes cheaper on short urban trips, and often more predictable because it doesn't surge as aggressively. However, availability is limited to the cities listed above, and wait times can stretch to 10–15 minutes during peak demand.
Tesla FSD vs. Waymo: Technology Stack Differences
The engineering philosophies behind Tesla FSD and Waymo are fundamentally opposed. Understanding these differences explains why their safety profiles and operational limits diverge.
Tesla FSD (Supervised) — Vision-Only
- Sensors: 8 cameras (surround-view), no lidar, no radar (removed in 2021+ models)
- Compute: Custom Tesla FSD chip (HW4 in current models), ~300+ TOPS
- Approach: End-to-end neural network trained on billions of miles of fleet video data
- Mapping: Minimal reliance on HD maps; system interprets the environment in real time
- Geofencing: None — designed to work on any road (but driver remains responsible)
Waymo (6th-Gen Driver) — Multi-Sensor Fusion
- Sensors: Lidar (multiple units, 360° coverage), cameras, radar, external audio sensors
- Compute: Custom Waymo compute platform
- Approach: Sensor fusion with HD mapping; system cross-references real-time perception against pre-mapped environment
- Mapping: Centimeter-precision HD maps of every operational street, updated continuously
- Geofencing: Strict — vehicle will not operate outside mapped/approved zones
Tesla's vision-only approach is more scalable (cameras are cheap; lidar is not), but it places enormous demands on software to handle edge cases that lidar's precise depth-sensing resolves more reliably — particularly in low-light, heavy rain, or scenarios with partial occlusion. Waymo's multi-sensor stack is more expensive per vehicle but provides redundant perception layers.
Regulatory Landscape: Who Approves What?
The regulation of autonomous driving in the U.S. is a patchwork of federal and state authority:
- Federal (NHTSA): Sets vehicle safety standards and investigates defects. NHTSA has active investigations into Tesla Autopilot/FSD related to crashes near emergency scenes and the system's driver-monitoring adequacy. NHTSA does not "approve" Level 2 systems before sale — they are self-certified by manufacturers.
- State level: States control whether driverless (Level 4) operations are permitted on public roads. California's DMV and Public Utilities Commission (CPUC) issue permits for autonomous vehicle testing and commercial deployment. Arizona has taken a permissive approach with minimal state-level barriers. Other states have varying frameworks.
- Level 2 systems (Tesla FSD, GM Super Cruise, Ford BlueCruise): Can be sold and used in all 50 states because they are classified as driver-assist, not autonomous. The driver is legally responsible.
- Level 4 robotaxis (Waymo, Zoox, Cruise): Require specific state and local permits. Cruise's operations were suspended in California in late 2023 following a serious incident in San Francisco; as of 2026, Cruise has relaunched in limited, phased operations with safety drivers before pursuing driverless service again.
Current Limitations: Where Both Systems Fall Short
Tesla FSD (Supervised) Limitations
- Weather: Heavy rain, snow, fog, and direct sun glare can degrade camera-only perception. System may disengage unexpectedly.
- Construction zones: Temporary lane shifts, cones, and flaggers remain challenging. FSD may not correctly interpret temporary traffic patterns.
- Unprotected left turns: Complex multi-lane intersections with oncoming traffic are a known edge case.
- Emergency scenes: NHTSA's investigation centers on FSD/Autopilot behavior near stationary emergency vehicles with flashing lights — a recurring crash pattern.
- Driver complacency: The system is good enough to inspire over-trust but not good enough to handle all scenarios. This is the core Level 2 paradox.
- No geofencing: FSD will attempt to drive on any road, including roads it may not handle well. The burden of judgment falls on the driver.
Waymo Robotaxi Limitations
- Geofencing: Service is only available in mapped zones within approved cities. No rural, highway, or cross-city trips outside boundaries.
- Weather: Waymo avoids operating in heavy rain, dense fog, and snow. Service may be suspended during severe weather events.
- Stalled-vehicle scenarios: Waymo vehicles have occasionally stopped in travel lanes when encountering ambiguous situations, requiring remote assistance.
- Wait times: Peak-hour availability can be limited; 10-15 minute waits are common.
- Highway driving: Waymo's urban service is primarily surface-street; limited freeway operation in Phoenix but not the primary use case.
- No personal vehicle ownership: You can't buy a Waymo. It's a ride-hail service only.
Is Tesla Self-Driving Safe? The Honest Answer
Tesla FSD (Supervised) is a sophisticated driver-assistance system that can reduce fatigue on long highway drives and handle some urban navigation. It is not self-driving, and treating it as such is dangerous. The safety data Tesla publishes is real but context-dependent — heavily weighted toward highway conditions and not independently verified with the rigor of a controlled study.
Waymo's Level 4 robotaxi has a stronger claim to the "self-driving" label and, based on available data from millions of driverless urban miles, demonstrates a safety record competitive with or better than human drivers in comparable conditions. But Waymo is a service, not a car you can own, and it's only available in a handful of cities.
If you're considering a Tesla with FSD, buy it for the driver-assistance capability and convenience — not because you believe the car will drive itself safely without your attention. You remain the driver, legally and practically.
Frequently Asked Questions
Is Tesla FSD really self-driving?
No. Tesla Full Self-Driving (Supervised) is an SAE Level 2 driver-assist system. Despite the name, you must remain engaged, keep your hands near the steering wheel, and monitor the road at all times. You are legally responsible for the vehicle's actions. Only Level 4 services like Waymo operate without a human driver.
Where can I use a robotaxi service right now?
Waymo One is commercially available in Phoenix, San Francisco, Los Angeles, and Austin as of early 2026. Zoox is testing in Las Vegas and Foster City, CA, but has not launched broad commercial service. Cruise is in limited phased relaunch. Tesla does not operate a robotaxi service — its FSD system requires a licensed driver in the vehicle.
How much does Waymo cost compared to Uber?
Waymo pricing is generally comparable to UberX, typically $10–$14 for a 5-mile urban trip. Waymo's pricing tends to be more predictable with less surge variability, but service is only available in select cities and wait times can be longer during peak hours.
How safe are robotaxis compared to human drivers?
Waymo reports that across over 10 million driverless trips, its vehicles have been involved in fewer police-reported crashes per million miles than human benchmarks in similar urban environments. However, robotaxis operate in geofenced, mapped areas and avoid severe weather — conditions that are inherently lower-risk than the full spectrum of human driving. Direct apples-to-apples comparisons remain difficult.
Can Tesla FSD drive me home if I'm tired or intoxicated?
No. Using FSD while impaired is illegal and extremely dangerous. FSD is a Level 2 system that requires an alert, sober, licensed driver ready to take over instantly. If you are too tired or impaired to drive, you are too impaired to supervise FSD. Use a rideshare or taxi instead.
What happens if a Tesla on FSD gets into a crash?
The driver is liable. Tesla's terms of service and the system's own warnings make clear that FSD (Supervised) is a driver-assist feature. You are responsible for monitoring the system and intervening when necessary. NHTSA may investigate systemic defects, but legal and insurance liability rests with the human driver.
Will Tesla ever have a true Level 4 robotaxi?
Tesla has announced plans for a robotaxi service and has demonstrated prototype vehicles. As of early 2026, Tesla does not operate a commercial Level 4 driverless service. Any future robotaxi would require regulatory approval from state authorities, and timelines have historically shifted. Treat announced capabilities as preliminary until commercially live.
