If you have seen a Waymo Jaguar I-PACE navigate downtown Phoenix without anyone behind the wheel, you have probably asked the same question millions of riders are asking: are self driving cars safe enough to trust with your daily commute?
The honest answer requires separating consumer driver-assist systems (which still require your full attention) from true Level 4 robotaxi services (which operate without a driver in geofenced zones). The safety data for the latter is substantial, but it is not without caveats. Here is what the numbers, regulators, and real-world miles actually show as of early 2026.
First: What "Self-Driving" Actually Means (SAE Levels Explained)
Before evaluating safety, we have to define the automation level. The SAE International J3016 standard defines six levels of driving automation, and conflating them is the single biggest source of public confusion.
| Level | Name | Who Drives? | Examples |
|---|---|---|---|
| 0–1 | No Automation / Driver Assist | Human (system may steer OR accelerate) | Cruise control, lane-keep assist |
| 2 | Partial Automation | Human (must monitor at all times) | Tesla Autopilot/FSD Supervised, GM Super Cruise, Ford BlueCruise |
| 3 | Conditional Automation | System (human must take over on request) | Mercedes DRIVE PILOT (limited markets/conditions) |
| 4 | High Automation | System (no human needed in geofence) | Waymo, Zoox, Cruise (where approved) |
| 5 | Full Automation | System (anywhere, any condition) | Does not exist commercially |
When this article discusses "self-driving cars" in the context of robotaxi safety data, it refers to SAE Level 4 services. Consumer vehicles with Level 2 systems like Tesla FSD (Supervised) are not self-driving — the driver remains legally and operationally responsible.
Where Can You Actually Use a Robotaxi Right Now?
As of early 2026, fully driverless Level 4 robotaxi services remain limited to specific cities and geofenced zones within those cities. You cannot hail one just anywhere. Here is the current operational landscape:
| Operator | Vehicle Platform | Live Cities | Hours |
|---|---|---|---|
| Waymo One | Jaguar I-PACE (custom 5th-gen sensor suite); Zeekr RT (announced) | Phoenix metro, San Francisco, Los Angeles, Austin | 24/7 in most zones |
| Cruise | Chevrolet Bolt EV (Origin paused post-2024 permit suspension) | Limited re-launch in Phoenix, Dallas, Houston (safety driver present in some zones) | Restricted hours |
| Zoox | Zoox purpose-built vehicle (no steering wheel) | Las Vegas (Strip corridor), Foster City, CA (limited) | Limited hours |
| Baidu Apollo Go | Various (China-specific platforms) | Wuhan, Beijing, Chongqing, Shenzhen (China) | Extended hours |
Key takeaway: If you live outside these metro areas, Level 4 robotaxi service is not yet available to you. Announced expansion plans (e.g., Waymo targeting additional cities) are not live service — always check the operator's app for your specific address before planning a trip.
How Safe Are Robotaxis? What the Real Data Shows
This is the core question, and the answer depends on what baseline you compare against. Let us look at the concrete data rather than marketing claims.
- Waymo cumulative miles (through 2025): Over 30 million fully driverless miles logged across Phoenix, SF, and LA, per Waymo's published safety reports.
- Waymo injury-causing crash rate: Waymo's published safety data claims an 85% reduction in injury-causing crash rates compared to a human benchmark, based on internal analysis of over 25 million driverless miles.
- NHTSA Standing General Order: The National Highway Traffic Safety Administration requires all ADS-equipped vehicles to report crashes. As of the latest reporting cycle, Level 4 vehicles were involved in hundreds of reported incidents — but the vast majority were low-speed, no-injury contacts (often rear-ended by human drivers at stops).
- Cruise (2023 suspension context): Cruise's California DMV permit was suspended in October 2023 after a pedestrian-dragging incident in San Francisco. The company has since restructured operations, re-introduced safety drivers in many zones, and is slowly re-expanding driverless service under heightened regulatory scrutiny.
- Disengagement rates: California DMV disengagement reports (miles between required human takeovers) are less relevant for Level 4 vehicles that have no driver to intervene, but historical data showed Waymo averaging over 17,000 miles between disengagements in testing — far ahead of most competitors.
Context Matters: What the Data Does and Doesn't Tell Us
The data above is encouraging but comes with important nuance:
- Geofenced conditions: Robotaxis operate in mapped, well-maintained urban environments with clear lane markings. They do not handle rural roads, unmapped construction zones, or severe weather well.
- Low-speed bias: Most robotaxi trips are urban, averaging 25–35 mph. The safety profile at 65+ mph highway speeds is less established.
- Rear-end vulnerability: A significant share of robotaxi incidents involve human-driven vehicles rear-ending the robotaxi at stops — the robotaxi brakes conservatively, and the human behind it does not. This is a "safe but overly cautious" problem, not a "the AI cannot drive" problem.
- Edge cases remain: Unusual scenarios — a person in a costume, a fallen tree branch, a police officer directing traffic with hand signals — can still confuse perception systems, leading to the vehicle stopping and requesting remote assistance.
Robotaxi Pricing: How Does It Compare to Uber and Lyft?
Pricing varies by city, time of day, and promotional pricing (operators often subsidize early rides to build ridership). Here is a representative comparison based on typical 2026 fares in Phoenix and San Francisco:
| Service | Base Fare | Per-Mile Rate | Typical 5-Mi Trip | Surge Pricing? |
|---|---|---|---|---|
| Waymo One | ~$5.00 | ~$1.50/mi | $12–$16 | Yes (demand-based) |
| UberX | ~$3.50 | ~$1.25/mi | $10–$14 | Yes |
| Lyft Standard | ~$3.50 | ~$1.20/mi | $9–$13 | Yes |
The reality: Robotaxi pricing is currently comparable to or slightly above standard rideshare. The long-term economic thesis is that removing the driver (who typically earns 60–75% of the fare) should make robotaxis significantly cheaper — but operators have not yet reached the scale and utilization to pass those savings fully to riders. Expect pricing to remain in the rideshare range through 2026, with promotional discounts in newer markets.
How to Use a Robotaxi Service
- Download the app: Waymo One (iOS/Android), Cruise app, or Zoox app depending on your city.
- Set pickup/dropoff: The app will confirm whether both locations are within the geofenced service area.
- Wait for the vehicle: Typical wait times are 3–8 minutes in well-served zones.
- Unlock and ride: The app unlocks the doors when the vehicle arrives. You buckle up, confirm your destination on the in-car screen, and the trip begins.
- Payment: Automatic via the app, same as rideshare. Tipping is not expected (no driver).
Under the Hood: How Robotaxis Actually "See" the Road
Understanding the technology stack helps explain both the capabilities and the limitations of current Level 4 systems.
| Sensor Type | Function | Range | Strengths / Weaknesses |
|---|---|---|---|
| LiDAR (roof-mounted, typically 360°) | Precise 3D mapping of surroundings | 150–250 meters | Excellent depth perception; expensive; performance can degrade in heavy rain/fog |
| Cameras (8–20+ around vehicle) | Traffic lights, signs, lane markings, object classification | 100–200 meters | Rich semantic data; affected by glare, darkness, occlusion |
| Radar (front/rear/corners) | Velocity detection, object tracking | 150–300 meters | Works in rain/fog; lower resolution than LiDAR |
| Ultrasonic | Close-range obstacle detection | 1–5 meters | Low-speed maneuvers, parking |
| HD Maps + GPS/IMU | Pre-built centimeter-accurate maps | Geofence-wide | Provides context; must be constantly updated for construction/changes |
LiDAR vs. camera-only: Waymo, Cruise, and Zoox all use LiDAR-heavy stacks. Tesla's consumer FSD (Supervised) system uses cameras only — a fundamentally different approach that remains Level 2 (driver-supervised) and is not used in any commercial robotaxi service.
The reliance on HD maps and geofencing is critical to understand: these vehicles are not "figuring out" new roads in real time the way a human does. They are executing within a pre-mapped, constantly updated digital environment. Leave the geofence, and the service simply will not operate.
The Regulatory Landscape: Who Approves Robotaxi Operations?
Robotaxi deployment in the United States is governed by a patchwork of state and local authorities — there is no single federal "robotaxi license."
- California (CPUC + DMV): The California Public Utilities Commission issues permits for commercial passenger service, while the DMV regulates testing. The CPUC suspended Cruise's driverless permit in October 2023 and has since granted phased re-entry approvals.
- Arizona: The most permissive major market. Waymo has operated commercially in the Phoenix metro since 2020 with relatively light regulatory friction. The state does not require a specific commercial robotaxi permit beyond standard business and vehicle registration requirements.
- Texas: Austin allows Waymo operations. Dallas and Houston have seen Cruise re-expansion with safety drivers. Texas passed HB 3026 (2023) establishing a framework for autonomous vehicle operations.
- Nevada: Zoox operates its purpose-built vehicle in Las Vegas under a state-issued testing and deployment framework.
- Federal (NHTSA): NHTSA does not "approve" robotaxi services but sets vehicle safety standards and requires crash reporting under its Standing General Order. NHTSA has also issued exemptions allowing vehicles without steering wheels (like the Zoox vehicle and GM Cruise Origin) to operate on public roads.
The regulatory environment remains fluid. Permits can be (and have been) suspended following safety incidents, and operators must continuously demonstrate safety performance to maintain authorization.
Current Limitations: What Robotaxis Cannot Do
- Geofence boundaries: Service is limited to pre-mapped zones. A trip from downtown Phoenix to Tucson? Not possible by robotaxi.
- Severe weather: Heavy rain, snow, fog, and dust storms degrade LiDAR and camera performance. Most services will pause operations or refuse rides in these conditions.
- Construction zones: While improving, unmapped lane closures, temporary barriers, and flaggers can still cause the vehicle to stop and request remote assistance, adding 5–15 minutes to your trip.
- Highway driving: Most current robotaxi services are optimized for surface streets at 25–45 mph. Limited highway operation exists (Waymo on certain Phoenix freeways) but is not the primary use case.
- Complex pickups: Gated communities, parking garages, private driveways, and crowded event venues can be difficult for the vehicle to navigate for pickup/dropoff.
- No rural or suburban coverage: If you live outside the urban core of a supported city, robotaxi service is unlikely to reach you.
- Accessibility: While Waymo and Zoox have made efforts to accommodate wheelchair users, not all vehicles in current fleets are fully ADA-accessible. Check the app for accessible vehicle options.
So — Are Self Driving Cars Safe Enough to Ride In?
Based on the available data from millions of driverless miles, Level 4 robotaxis in approved service areas have demonstrated a safety record that is competitive with or better than the human average for comparable urban driving conditions. Waymo's published data shows meaningful reductions in injury-causing crash rates, and the vast majority of reported incidents are low-severity contacts — often caused by other human-driven vehicles.
However, "safe" is not the same as "perfect" or "universally available." The technology excels within its operational design domain (mapped urban streets, fair weather, moderate speeds) and falls short outside it. The Cruise suspension of 2023 was a reminder that even established operators can encounter serious safety failures, and regulatory oversight is both necessary and active.
If you are in a supported city and your trip fits the service profile — urban, fair weather, within the geofence — the data supports that a robotaxi ride is a reasonable, data-backed transportation choice. If you need to drive on a snowy mountain road or navigate an unmapped rural route, you are firmly in human-driver territory.
Frequently Asked Questions
Is Tesla FSD considered a self-driving car?
No. Tesla's Full Self-Driving (Supervised) is an SAE Level 2 driver-assistance system. The driver must remain engaged, monitor the road, and be prepared to take over at any moment. It is not a robotaxi service, and Tesla does not operate a commercial driverless ride-hailing platform as of 2026. Calling it "self-driving" is marketing language, not a technical or legal classification.
How many miles have robotaxis driven without a driver?
Waymo alone has accumulated over 30 million fully driverless miles (no human in the driver's seat) across its Phoenix, San Francisco, and Los Angeles operations through late 2025. Combined with Cruise, Zoox, and Baidu Apollo Go, the global total of Level 4 driverless miles likely exceeds 60 million — a substantial dataset, though still a fraction of the ~3 trillion miles driven annually by humans in the US alone.
Can a robotaxi get into a serious accident?
Yes. No transportation mode is risk-free. NHTSA's crash reporting database includes incidents involving Level 4 vehicles, though the majority are low-speed, no-injury events. The Cruise pedestrian incident in San Francisco (October 2023) demonstrated that serious incidents can and do occur, and they carry significant regulatory consequences.
What happens if the robotaxi gets confused or stuck?
The vehicle will typically come to a safe stop and contact a remote assistance operator — a human who can view the vehicle's sensor data and provide guidance (e.g., "proceed around the obstacle" or "reroute"). This adds time to your trip but is a designed safety fallback, not a system failure.
Will robotaxis replace Uber and Lyft drivers?
Not in the near term. Robotaxi availability is limited to a handful of cities, and scaling requires massive capital investment in vehicles, mapping, remote assistance infrastructure, and regulatory approvals. Uber and Lyft have partnered with robotaxi operators (Uber integrates Waymo rides in Phoenix) rather than being replaced by them. A hybrid model — robotaxis for routine urban trips, human drivers for complex or out-of-area rides — is the likely medium-term outcome.
Are robotaxis available in my city?
As of early 2026, fully driverless Level 4 robotaxi service in the US is available in Phoenix, San Francisco, Los Angeles, Austin, and limited zones in Las Vegas. Check the Waymo One, Cruise, or Zoox app with your specific address to confirm coverage. Announced future expansions are not yet live.
