Every few months, a headline about a Tesla autonomous car crash reignites the debate: how safe is Tesla's Full Self-Driving (FSD) technology, and how does it compare to purpose-built robotaxi services like Waymo? The answer requires separating marketing language from engineering reality — and understanding the fundamental gap between Level 2 driver-assist and Level 4 driverless systems.

This analysis examines the available crash data, regulatory filings, and operational constraints shaping autonomous vehicle safety in 2026.

Understanding the Autonomy Gap: Level 2 vs. Level 4

The most critical distinction in autonomous vehicle safety is the SAE automation level. Tesla's FSD (Supervised) and Autopilot are SAE Level 2 systems — advanced driver-assistance that requires continuous human supervision. Waymo's robotaxi service operates at SAE Level 4 — fully driverless within defined operational domains.

SAE Automation Levels: What They Actually Mean
Level Who Drives System Type Examples (2026)
Level 2 Human (must monitor constantly) Driver-assist Tesla FSD Supervised, GM Super Cruise, Ford BlueCruise
Level 3 System (human on standby) Conditional automation Mercedes Drive Pilot (limited markets)
Level 4 System (no human needed in zone) High automation Waymo One, Zoox (limited geofenced areas)
Level 5 System (anywhere, any condition) Full automation None commercially available
Key Point: When a Tesla FSD crash occurs, the driver bears legal responsibility. When a Waymo robotaxi crashes, Waymo bears liability. This is not a semantic difference — it defines who is accountable and how safety systems are designed.

Tesla FSD Crash Data: What We Know

Tesla does not publish granular crash data specific to FSD-engaged miles. Instead, the company releases quarterly safety reports showing aggregate accident rates per million miles driven with Autopilot/FSD active versus inactive. These reports have faced scrutiny from the National Highway Traffic Safety Administration (NHTSA) for methodology concerns.

NHTSA Investigation and Standing General Order

Under NHTSA's Standing General Order (SGO), manufacturers must report crashes involving Level 2+ and higher automation systems. Tesla has reported hundreds of incidents since the order took effect. Notable patterns include:

  • Rear-end collisions: FSD-engaged vehicles striking stationary or slow-moving traffic, particularly in low-light conditions or construction zones.
  • Lane departure: Vehicles drifting into adjacent lanes or off-road when lane markings are faded or obscured.
  • Emergency vehicle encounters: Multiple crashes involving Tesla vehicles striking parked or slow-moving first-responder vehicles with flashing lights — a scenario NHTSA specifically investigated.
Data Limitation: Tesla's safety reports show crashes per million miles with Autopilot engaged, but do not isolate FSD Supervised usage or control for road type (highway vs. urban). Direct comparison to Waymo's per-mile incident rate requires caution.

Waymo Safety Record: The Level 4 Baseline

Waymo publishes more granular safety data, partly because its driverless model requires demonstrating safety to regulators before expansion. As of early 2026, Waymo has completed over 30 million fully driverless miles across its operational cities.

Waymo Safety Data (as reported 2025-2026)
Metric Waymo Driverless Human Baseline (U.S. avg)
Fatal crashes per billion miles 0 (as of Q1 2026) ~12-15
Injury-causing crashes per million miles ~0.4 ~1.1
Police-reported crashes per million miles ~3.5 ~4.2

Waymo's data, reviewed by independent researchers and submitted to NHTSA's AV safety portal, suggests lower crash rates than human drivers in comparable urban environments. However, Waymo operates under strict geofencing, avoids highways, and pulls over in adverse weather — conditions that do not reflect the full spectrum of driving scenarios Tesla FSD encounters.

Technology Stack: Camera-Only vs. Sensor Fusion

The engineering philosophies behind Tesla FSD and Waymo's robotaxi platform diverge sharply, with direct safety implications.

Tesla: Vision-Only Approach

Tesla removed radar and ultrasonic sensors from its vehicles starting in 2021-2022, relying exclusively on a camera suite (typically 8 cameras providing 360-degree coverage) paired with neural network processing. The argument: human drivers use only vision, so cameras should suffice.

Limitations: Camera-only systems struggle in heavy rain, fog, direct sun glare, and low-light scenarios. Depth perception relies on software inference rather than direct measurement.

Waymo: Lidar + Camera + Radar Fusion

Waymo's fifth-generation Driver system combines:

  • Lidar: 360-degree laser scanning for precise 3D mapping and object detection, effective in darkness and most weather.
  • Cameras: High-resolution visual data for object classification (pedestrian vs. cyclist, traffic light state).
  • Radar: Velocity measurement and redundancy in heavy precipitation.

This sensor fusion approach provides redundancy — if one sensor type degrades (camera blinded by sun, lidar scattered by fog), others compensate. The trade-off is cost: Waymo's sensor suite adds tens of thousands of dollars per vehicle, viable only for fleet operations.

Where Robotaxis Operate Today (2026)

Level 4 robotaxi services remain geographically constrained. Here's where you can actually hail a driverless ride:

Robotaxi Service Availability (Q1 2026)
Service Cities Operating Hours Access
Waymo One Phoenix, San Francisco, Los Angeles, Austin (limited) 24/7 in most zones Public app (waitlists in some markets)
Zoox Las Vegas (limited zone), Foster City, CA (testing) Limited hours Invite-only / employee testing
Cruise Suspended operations (as of early 2026) N/A N/A

Tesla's position: Tesla does not operate a robotaxi service. FSD Supervised is available to Tesla owners nationwide but requires an attentive driver at all times. Tesla's announced "Robotaxi" platform (unveiled as a concept) has no confirmed launch date or regulatory approval as of 2026.

Pricing: Robotaxi vs. Rideshare

Where Waymo One operates, pricing is competitive with Uber/Lyft but varies by market and time:

Typical Ride Pricing (2026, select markets)
Service Base Fare Per Mile Per Minute
Waymo One (Phoenix) $3-5 $1.20-1.50 $0.30-0.40
UberX (Phoenix avg) $2-4 $1.00-1.30 $0.25-0.35
Lyft (Phoenix avg) $2-4 $1.00-1.25 $0.25-0.35

Waymo pricing occasionally undercuts rideshare during off-peak hours but can exceed it during high demand (Waymo does not use surge pricing, but base rates vary). A 5-mile trip in Phoenix typically costs $10-15 on Waymo versus $9-14 on Uber/Lyft.

Current Limitations: What Robotaxis Cannot Do

Level 4 robotaxis operate under strict constraints that human drivers and Level 2 systems do not face:

  • Geofencing: Waymo vehicles will not enter unmapped zones or areas outside approved service boundaries.
  • Weather restrictions: Heavy rain, snow, or fog triggers pull-over protocols. Waymo does not operate in snow-belt cities.
  • No highway driving: Current Waymo operations avoid freeways (speed limits above 45 mph in most zones).
  • Construction zones: Unmapped detours or lane closures can cause the vehicle to stop and request remote assistance.
  • Pickup/dropoff precision: Vehicles may stop 50-100 feet from requested addresses if the exact spot is obstructed or unmapped.

Tesla FSD Supervised, as a Level 2 system, does not have these constraints — but the human driver must compensate for the system's limitations in real-time.

Regulatory Landscape: Who Approves What

Autonomous vehicle regulation in the U.S. operates at both federal and state levels:

  • Federal (NHTSA): Sets vehicle safety standards, investigates defects, and collects crash data via the Standing General Order. NHTSA does not "approve" autonomous systems but can order recalls.
  • State (DMV/CPUC equivalents): California, Arizona, Texas, and Nevada have specific permitting processes for driverless testing and commercial deployment. Waymo holds permits in multiple states; Tesla's FSD does not require special permits because it is classified as driver-assist.
  • Local (city/county): Some municipalities negotiate operating agreements or restrict robotaxi zones (San Francisco initially limited Waymo hours and zones).

The Insurance Institute for Highway Safety (IIHS) has called for standardized testing protocols for Level 2+ systems, noting that current federal regulations do not address partial automation safety comprehensively.

Frequently Asked Questions

Is Tesla FSD actually self-driving?

No. Tesla Full Self-Driving (Supervised) is an SAE Level 2 driver-assist system. The driver must remain attentive, hands on or near the wheel, and ready to take over instantly. The "Supervised" label, added in 2024, reflects this requirement.

How many Tesla FSD crashes have been reported?

Under NHTSA's Standing General Order, Tesla has reported hundreds of crashes involving Autopilot/FSD since 2021. Exact FSD-specific counts are not publicly broken out from Autopilot data. NHTSA has opened multiple investigations into Tesla's driver monitoring and crash patterns.

Are Waymo robotaxis safer than human drivers?

Waymo's published data suggests lower crash rates per mile than U.S. human driver averages, particularly for injury-causing incidents. However, Waymo operates in constrained environments (no highways, no severe weather, geofenced zones), making direct comparison complex.

Can I use a robotaxi in my city?

As of 2026, Waymo One operates in Phoenix, San Francisco, Los Angeles, and limited Austin zones. Zoox operates a small invite-only service in Las Vegas. Most U.S. cities do not have robotaxi access.

Will Tesla launch a driverless robotaxi service?

Tesla has announced plans for a "Robotaxi" platform but has not received regulatory approval for driverless operation, nor confirmed a launch date. Any such service would require Level 4 certification and state permits — a process Waymo spent over a decade navigating.