Tesla's Full Self-Driving (FSD) system has logged billions of miles on public roads, with California serving as both its home market and its most scrutinized testing ground. But parsing what "autonomous car testing" actually means for Tesla — versus what riders experience in a Waymo robotaxi — requires separating marketing language from regulatory reality and hard data.
This guide breaks down exactly where Tesla's autonomous driving technology stands in California as of 2026, what the safety numbers say, how the technology stack compares to purpose-built robotaxis, and what buyers and riders should realistically expect.
SAE Automation Levels: Where Tesla FSD Actually Sits
The single most important fact about Tesla's autonomous car testing in California is the SAE automation level. Despite the "Full Self-Driving" name, Tesla FSD Supervised remains a SAE Level 2 driver-assistance system. The driver must remain attentive, hands available, and ready to take over at any moment.
| Level | Name | Driver Role | Examples |
|---|---|---|---|
| Level 0 | No Automation | Full control | Basic cruise control |
| Level 1 | Driver Assistance | Hands on wheel, full attention | Lane-keep OR adaptive cruise |
| Level 2 | Partial Automation | Must supervise constantly | Tesla FSD, GM Super Cruise, Ford BlueCruise |
| Level 3 | Conditional Automation | Eyes off in specific conditions | Mercedes Drive Pilot (limited markets) |
| Level 4 | High Automation | No driver needed in geofence | Waymo One, Zoox |
| Level 5 | Full Automation | No driver, any condition | Does not exist commercially |
This distinction is not semantic. Level 2 means the human is legally and operationally the driver. Level 4 — what Waymo operates in San Francisco, Phoenix, and Los Angeles — means no human driver is present or needed within the service area. Tesla has not achieved, nor received regulatory approval for, Level 4 operations anywhere.
Tesla FSD Testing Footprint in California
Tesla's approach to autonomous car testing differs fundamentally from Waymo or Cruise. Rather than deploying a geofenced fleet of purpose-built robotaxis, Tesla runs its FSD software across its entire consumer fleet — every Model 3, Model Y, Model S, and Model X equipped with the FSD package becomes a test vehicle.
In California alone, Tesla has hundreds of thousands of FSD-equipped vehicles on the road. The state's DMV disengagement reports, which manufacturers operating autonomous test programs must file annually, provide one window into system performance. However, Tesla's consumer FSD fleet operates under different regulatory categories than dedicated test fleets with safety drivers, which complicates direct comparison.
What Tesla's California Testing Looks Like in Practice
- Consumer fleet testing: Everyday Tesla owners using FSD Supervised on commutes, errands, and road trips throughout the state
- Employee testing: Tesla employees running pre-release FSD builds in the Bay Area and greater California
- Robotaxi prototype testing: Tesla's announced "Cybercab" prototype has been spotted testing on closed courses and limited public roads, but no commercial robotaxi service is operational
| Service / System | SAE Level | California Availability | Rider Access |
|---|---|---|---|
| Tesla FSD Supervised | Level 2 | Statewide (owner vehicles) | Tesla owners only; not a ride-hail service |
| Tesla Cybercab (announced) | Targeting Level 4 | Not yet approved or launched | No public access |
| Waymo One | Level 4 | SF, LA, Phoenix, Austin | Open to public via Waymo app |
| Zoox | Level 4 | Limited SF/Las Vegas testing | Invite-only / employee rides |
The Technology Stack: Camera-Only vs. Lidar
Tesla's autonomous driving stack is philosophically distinct from virtually every other player in the space. Understanding this difference is essential for evaluating what Tesla's California testing data actually represents.
Tesla removed all radar and ultrasonic sensors from new vehicles starting in 2022–2023, relying exclusively on a camera suite (typically 8 cameras providing 360° coverage) paired with neural-network-based perception. The system processes visual data through Tesla's custom HW4 (Hardware 4) computer, running increasingly sophisticated end-to-end neural networks trained on billions of miles of fleet driving data.
How this differs from Waymo/Zoox: Purpose-built robotaxis use a redundant sensor suite including lidar (light detection and ranging), multiple radars, and cameras. Lidar provides precise 3D depth mapping independent of lighting conditions — a capability Tesla's vision-only system must approximate through software.
Trade-Offs of Camera-Only Perception
| Factor | Tesla (Camera-Only) | Waymo (Lidar + Radar + Camera) |
|---|---|---|
| Hardware cost per vehicle | ~$2,000–$3,000 (cameras + compute) | Estimated $50,000–$100,000+ (lidar stack) |
| Low-light / night performance | Degraded; reliant on ambient/IR illumination | Strong; lidar is lighting-independent |
| Depth accuracy | Software-estimated (neural net inference) | Centimeter-precise (direct lidar measurement) |
| Scalability to millions of cars | High — already deployed at scale | Low — cost and calibration constraints |
| Weather sensitivity | Heavy rain, fog, sun glare impair cameras | Lidar affected by heavy precipitation; radar more resilient |
Tesla's bet is that neural networks trained on massive real-world datasets will eventually match or exceed lidar-based perception at a fraction of the cost. As of 2026, that bet remains unproven at the Level 4 threshold required for driverless operation.
Safety Data: What the Numbers Show
Evaluating the safety of Tesla autonomous car testing in California requires navigating a fragmented data landscape. Unlike Waymo, which publishes detailed safety reports and operates under California DMV autonomous vehicle permits with mandatory disengagement reporting, Tesla's FSD data comes primarily from the company's own quarterly safety reports and third-party analyses.
- Cumulative FSD miles: Tesla reported surpassing 3 billion cumulative FSD (Supervised) miles driven by early 2026, across all markets (not California-specific)
- Tesla's self-reported crash rate: Tesla's quarterly safety reports claim approximately 1 crash per 4–5 million miles driven with FSD active, versus the U.S. average of ~1 crash per 670,000 miles (per NHTSA national data). However, these figures are not independently verified and exclude many incident categories
- NHTSA investigations: The National Highway Traffic Safety Administration has opened multiple investigations into Tesla's Autopilot and FSD systems, including crashes involving stationary emergency vehicles and the adequacy of driver-monitoring systems
- California DMV disengagement data: Companies with autonomous vehicle test permits (Waymo, Cruise, Zoox) report disengagements — moments when the system hands control back to a safety driver or the driver intervenes. Tesla's consumer FSD fleet is not subject to the same reporting requirements, making apples-to-apples comparison difficult
Comparing Safety Records: Tesla FSD vs. Waymo
Waymo has published more granular safety data than any robotaxi operator. In its safety reports, Waymo documented over 7.1 million driverless miles in California alone, with a significant reduction in serious-injury-involved crashes compared to human benchmarks in comparable urban environments.
| Metric | Tesla FSD Supervised | Waymo One (Driverless) |
|---|---|---|
| SAE Level | Level 2 (driver required) | Level 4 (no driver) |
| Cumulative miles (approx.) | 3+ billion (all FSD, global) | 10+ million driverless (US) |
| Third-party safety audit | Not published | Published; peer-reviewed methodology |
| Driver present? | Yes — always required | No — fully driverless in service area |
| NHTSA standing investigation | Yes (multiple active probes) | No standing fleet-wide investigation |
The comparison is inherently asymmetric: Tesla's billions of miles are driven by consumer owners with a human safety net, while Waymo's millions are fully driverless. A system that performs well with an attentive human backup is not equivalent to one that must handle every scenario independently.
Regulatory Landscape in California
California has the most structured autonomous vehicle regulatory framework in the United States, governed primarily by the California DMV and the California Public Utilities Commission (CPUC).
Key Regulatory Bodies and Their Roles
- California DMV: Issues autonomous vehicle testing permits (with and without safety drivers) and deployment permits. Requires annual disengagement reports from permitted companies.
- CPUC: Regulates commercial passenger service — any company charging fares for autonomous rides must receive CPUC approval. Waymo holds a CPUC permit for commercial robotaxi service in San Francisco and Los Angeles.
- NHTSA (federal): Sets vehicle safety standards and investigates defects. Does not currently certify autonomous driving systems but can issue recalls and standing investigations.
Tesla's FSD operates under a different regulatory category than dedicated robotaxi test programs. Because FSD is classified as a Level 2 driver-assist feature (like adaptive cruise control or lane-keeping), it does not require a DMV autonomous vehicle testing permit. The human driver is legally responsible for the vehicle at all times. This means Tesla's "testing" happens under consumer product regulations, not the autonomous vehicle permit framework that governs Waymo and Cruise.
For Tesla to launch a true robotaxi service in California — where passengers ride without a driver — the company would need both a DMV deployment permit and a CPUC commercial passenger service permit. As of 2026, Tesla holds neither.
Current Limitations of Tesla FSD in California
- No geofence, but variable capability: Unlike Waymo's precisely mapped service areas, FSD operates anywhere but with highly variable performance. Unmapped construction zones, temporary road closures, and poorly marked rural roads remain challenging.
- Weather degradation: Heavy rain, fog (common in Northern California coastal areas and the Central Valley), and direct low-angle sun can impair camera-based perception, leading to reduced confidence or system disengagement.
- Edge cases: Unprotected left turns across heavy traffic, complex multi-lane roundabouts, emergency vehicle interactions, and unusual road users (e.g., horse-drawn carriages, oversized loads) remain scenarios where FSD may hesitate, make unexpected maneuvers, or require driver takeover.
- Driver monitoring gaps: Tesla uses a cabin camera to monitor driver attention, but the system has drawn criticism (and NHTSA scrutiny) for being less robust than the steering-wheel torque sensors used by GM's Super Cruise or Ford's BlueCruise.
- No Level 4 pathway approved: Tesla has not submitted an application for driverless testing or deployment permits in California. The Cybercab concept remains announced, not operational.
- Recall history: Tesla has issued multiple over-the-air recalls related to Autopilot/FSD, including a major December 2023 recall affecting over 2 million vehicles to improve driver attentiveness monitoring, with subsequent updates through 2024–2025.
Is a Tesla Robotaxi Service Coming to California?
Tesla CEO Elon Musk has repeatedly announced timelines for a Tesla robotaxi service, with the "Cybercab" unveil occurring in late 2024 and production targeted for 2026–2027. However, several barriers remain between announcement and a California rider hailing a driverless Tesla:
- Regulatory approval: Tesla must apply for and receive DMV and CPUC permits — a process that takes months to years, as Waymo's California expansion has demonstrated.
- Technology validation: Moving from Level 2 (with a human safety net) to Level 4 (no human) requires demonstrating that the system can handle the full spectrum of driving scenarios without intervention. Tesla's camera-only approach must prove this without lidar redundancy.
- Insurance and liability: In a Level 4 robotaxi, the operator (not the passenger) bears liability. Tesla would need to establish commercial insurance frameworks for a driverless fleet.
- Manufacturing scale: The Cybercab is a new vehicle platform. Even if approved, ramping production to meaningful fleet volumes takes time.
Bottom line: No Tesla robotaxi service is available to riders in California or anywhere else as of 2026. FSD Supervised remains a driver-assist tool for Tesla owners — capable and continually improving, but not autonomous.
Pricing Context: FSD Cost vs. Robotaxi Rides
Since Tesla doesn't operate a robotaxi service, there's no per-ride price to compare with Uber or Waymo. But understanding the cost structure of Tesla's autonomous driving offering provides useful context:
| Cost Item | Tesla FSD Supervised | Waymo One (where available) |
|---|---|---|
| Access model | Purchase ($8,000 one-time) or subscription (~$99/month) | Per-ride, via Waymo app |
| Typical ride cost | N/A (you own the car) | Comparable to Uber/Lyft; varies by distance and demand |
| Vehicle cost | Model 3 starts ~$38,990; Model Y ~$44,990 (2026 MSRP) | No vehicle purchase required |
| Driver needed? | Yes — you drive/supervise | No — fully driverless |
For a Tesla owner, FSD adds capability to a car they already own. For someone seeking autonomous transportation without ownership, Waymo is the only Level 4 option currently available in parts of California — and its pricing generally tracks with conventional rideshare services.
Frequently Asked Questions
Where can I use a Tesla autonomous car service right now?
You can't — not as a rider. Tesla does not operate any robotaxi or ride-hail service. Tesla FSD Supervised is available only to Tesla vehicle owners who purchase or subscribe to the FSD package. It works on roads statewide in California, but the owner must be in the driver's seat and remain attentive at all times.
How much does Tesla FSD cost vs. using Uber or Waymo?
Tesla FSD costs $8,000 as a one-time purchase or approximately $99/month as a subscription, on top of the vehicle purchase price. This is not a per-ride cost — it's a vehicle feature. A Waymo ride in San Francisco or Los Angeles typically costs in line with Uber/Lyft pricing for the same route, with no vehicle ownership required.
How safe is Tesla FSD based on real data?
Tesla self-reports that FSD-active vehicles experience roughly 1 crash per 4–5 million miles, which would be significantly better than the national average. However, these figures are not independently audited, exclude many incident types, and reflect a system with a human driver as a safety net. NHTSA has active investigations into FSD-related incidents. The safety profile is fundamentally different from a Level 4 robotaxi that must operate without any human backup.
What are the current limitations of Tesla FSD?
Key limitations include degraded performance in heavy rain, fog, and glare; difficulty with unmapped construction zones and complex intersections; variable handling of emergency vehicles and unusual road users; and a driver-monitoring system that has drawn regulatory scrutiny. FSD is not approved for driverless operation anywhere.
Is Tesla FSD truly self-driving?
No. Tesla FSD Supervised is a SAE Level 2 driver-assistance system. The word "Supervised" in the product name is intentional and legally required: the driver must remain fully engaged, monitor the road, and be prepared to take over instantly. Calling it "self-driving" is inaccurate and potentially dangerous. Only Level 4 systems like Waymo One operate without a human driver — and only within their approved geofenced service areas.
When will Tesla launch a robotaxi service in California?
No confirmed date exists. Tesla has announced the Cybercab concept and targeted production for 2026–2027, but launching a commercial driverless service in California requires DMV and CPUC permits that Tesla has not yet applied for. Regulatory approval timelines are measured in years, not months, based on Waymo's experience. Treat any announced timeline with skepticism until permits are formally granted.
