Where Robotaxis Actually Operate Today
Autonomous vehicle safety is best understood not through marketing claims but through operational data—miles driven, disengagement rates, incident reports, and the regulatory frameworks that govern where these vehicles can and cannot go. As of early 2026, fully driverless commercial robotaxi services in the United States remain limited to a handful of cities, operated primarily by Waymo (an Alphabet subsidiary) and, in restricted form, Amazon's Zoox. Cruise, once a major player, suspended all driverless operations in late 2023 following a serious incident in San Francisco and has not resumed full commercial service as of this writing.
The distinction between what's available to the public and what's still in testing is critical. Below is a breakdown of current commercial service areas where any rider can hail a driverless vehicle.
| Operator | City | Service Type | Hours | Vehicle |
|---|---|---|---|---|
| Waymo | Phoenix, AZ (metro-wide) | Commercial, driverless | 24/7 | Jaguar I-PACE (Gen 6 sensor suite) |
| Waymo | San Francisco, CA | Commercial, driverless | 24/7 | Jaguar I-PACE (Gen 6 sensor suite) |
| Waymo | Los Angeles, CA | Commercial, driverless | Limited hours, expanding | Jaguar I-PACE (Gen 6 sensor suite) |
| Waymo | Austin, TX | Commercial, driverless (launched late 2025) | Expanding | Jaguar I-PACE (Gen 6 sensor suite) |
| Zoox | Las Vegas, NV (limited corridor) | Commercial, driverless | Limited hours | Zoox purpose-built vehicle |
| Zoox | Foster City, CA | Testing / limited employee rides | Restricted | Zoox purpose-built vehicle |
Every other "autonomous" ride you might experience—whether in a Tesla with FSD Supervised, a GM vehicle with Super Cruise, or a Ford with BlueCruise—is a Level 2 driver-assist system that requires an attentive human driver at all times. That distinction is not semantic; it defines the entire safety conversation.
Understanding SAE Automation Levels: Why Level 4 Matters
The SAE International J3016 standard defines six levels of driving automation. Robotaxis like Waymo and Zoox operate at SAE Level 4—high automation within a defined operational design domain (ODD). No human driver is present or needed within that domain. Consumer systems like Tesla Autopilot, FSD Supervised, GM Super Cruise, and Ford BlueCruise are SAE Level 2—partial automation that requires continuous driver supervision. There is currently no Level 5 (full automation everywhere, all conditions) system available commercially anywhere in the world.
| Level | Name | Who Drives? | Examples |
|---|---|---|---|
| 0 | No Automation | Human | Any conventional car |
| 1 | Driver Assistance | Human (system assists steering OR speed) | Adaptive cruise control, lane-keeping assist |
| 2 | Partial Automation | Human (must supervise 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, approved in NV and CA) |
| 4 | High Automation | System (within defined ODD, no human needed) | Waymo, Zoox, Cruise (suspended) |
| 5 | Full Automation | System (all conditions, all locations) | None currently available |
This table is essential context for autonomous vehicle safety discussions. When a Level 2 system is involved in a crash, the driver bears responsibility because the system explicitly requires supervision. When a Level 4 robotaxi is involved in a crash, the operator (Waymo, Zoox) bears responsibility because no driver exists to intervene. The regulatory, insurance, and liability implications are entirely different.
Autonomous Vehicle Safety: What the Data Actually Shows
Waymo reported in 2024 that across over 7.1 million fully autonomous miles driven in Phoenix, San Francisco, and Los Angeles, its vehicles were involved in 76% fewer injury-causing crashes and 73% fewer property-damage-only crashes compared to human-driven benchmark data from the same cities. These figures were published in a peer-reviewed analysis and submitted to NHTSA as part of the agency's Standing General Order requiring crash reporting for ADS-equipped vehicles.
Let's unpack what those numbers mean—and what they don't.
Miles Driven and Intervention Rates
Waymo's fleet of Jaguar I-PACE vehicles, equipped with the company's 6th-generation sensor suite, had accumulated over 20 million fully driverless miles by late 2025. The California DMV requires annual disengagement reports from companies testing autonomous vehicles in the state. In the most recent reporting period, Waymo's disengagement rate (instances where a remote operator or the system itself flagged a scenario requiring human input) remained among the lowest in the industry, though direct comparison between companies is complicated by differing definitions of what constitutes a "disengagement."
Zoox, operating its purpose-built bidirectional vehicle (which has no steering wheel or pedals), has logged fewer public autonomous miles. The company has focused on controlled corridor operations in Las Vegas and testing in Foster City, California. Its safety data remains less publicly available than Waymo's, though the vehicle's design—redundant systems, 360-degree sensor coverage, and a maximum speed initially limited to 35 mph—reflects a conservative operational approach.
Incident Context and NHTSA Reporting
NHTSA's Standing General Order, issued in 2021 and updated subsequently, requires all manufacturers and operators of Level 2 and higher automated driving systems to report crashes. This has created the most comprehensive federal dataset on autonomous vehicle incidents to date. According to NHTSA's ADS crash reporting portal, the vast majority of reported incidents involving Level 4 vehicles have been low-speed, minor-severity events—often rear-end collisions where the robotaxi was struck by a human-driven vehicle, not the initiating party.
The Cruise suspension in October 2023, however, illustrates how a single high-severity event can reshape the industry. A Cruise vehicle in San Francisco struck a pedestrian who had been thrown into its path by a human-driven car, then dragged the pedestrian approximately 20 feet while attempting to pull over. The incident led to Cruise's California DMV permit suspension, a voluntary nationwide pause, and a significant restructuring of the company's operations. It underscored that autonomous vehicle safety is not just about crash rates—it's about post-collision behavior, edge-case handling, and the transparency with which operators communicate with regulators.
How Robotaxi Safety Compares to Human Drivers
The human benchmark is sobering. According to NHTSA data, approximately 42,795 people died in U.S. motor vehicle crashes in 2023, a rate of roughly 1.27 fatalities per 100 million vehicle miles traveled. Robotaxis have not recorded a pedestrian fatality in commercial driverless operation as of early 2026. However, the sample size remains small: even 20 million autonomous miles is a fraction of the roughly 3.2 trillion miles driven annually by U.S. motorists. Statistically meaningful fatality-rate comparisons will require hundreds of millions of autonomous miles across diverse environments.
Robotaxi Technology Stack: How These Vehicles See the Road
Level 4 robotaxis rely on sensor fusion—combining data from multiple sensor types to build a real-time 3D model of the vehicle's surroundings. This is fundamentally different from Tesla's camera-only approach used in FSD Supervised (a Level 2 system). The redundancy in sensor types is a deliberate safety architecture choice for driverless operation.
| Sensor Type | Waymo (Gen 6, Jaguar I-PACE) | Zoox (Purpose-Built Vehicle) |
|---|---|---|
| LiDAR | Custom-designed roof-mounted + peripheral units; range ~500m | Multiple roof and corner units; 360° coverage |
| Cameras | 29+ cameras (narrow, wide, forward, peripheral) | Multi-camera array, all-around coverage |
| Radar | 6 radar units (long-range + short-range) | Multiple radar units |
| Audio sensors | External microphones (detect sirens, horns) | External audio detection |
| HD Maps | Pre-mapped geofenced areas; centimeter-level precision | Pre-mapped corridors |
| Redundancy | Dual-compute, redundant braking and steering | Fully redundant systems; bidirectional drive |
LiDAR (Light Detection and Ranging) is the defining sensor for most Level 4 operators. It emits laser pulses to create a precise, real-time 3D point cloud of the environment, functioning reliably in darkness and most lighting conditions. Its limitations include degraded performance in heavy rain, snow, and dense fog—conditions that scatter laser light. Radar excels at measuring the velocity and distance of objects and performs well in adverse weather but offers lower resolution. Cameras provide rich visual data for object classification (distinguishing a pedestrian from a mailbox, reading traffic signals) but struggle in low-light or high-glare conditions.
The combination creates overlapping coverage: if one sensor type is degraded, others compensate. This redundancy is non-negotiable for Level 4 operation where no human driver exists as a fallback. Tesla's camera-only FSD approach, by contrast, relies on neural networks processing visual data from eight cameras—a design choice that remains controversial in the autonomous vehicle safety community precisely because it lacks the sensor redundancy that most safety engineers consider essential for driverless operation.
Geofencing: The Invisible Boundary
Every current Level 4 robotaxi operates within a geofenced operational design domain (ODD)—a precisely mapped area where the system has been validated to operate. Waymo's Phoenix geofence covers much of the metro area, including suburban streets, multi-lane arterials, and highway segments. Its San Francisco geofence encompasses most of the city but initially excluded certain highway on-ramps and neighborhoods with steep grades. These boundaries are not arbitrary; they reflect where the operator has accumulated sufficient testing data and where HD maps have been built and validated.
If a rider requests a destination outside the geofence, the app simply will not allow the trip. The vehicle cannot drive itself beyond its mapped domain.
Pricing: Robotaxi vs. Uber and Lyft
Robotaxi pricing in 2026 is designed to be competitive with conventional rideshare, though it varies by city, time of day, and trip distance. Waymo charges through its Waymo One app, and pricing generally falls within the range of UberX and Lyft standard fares, sometimes slightly higher during peak demand periods but without the surge multipliers that human-driven rideshare can apply.
| Service | Typical Fare (5 mi) | Surge Pricing | Tipping |
|---|---|---|---|
| Waymo One | $12–$18 | No surge; may have wait times during peak | No tip (no driver) |
| UberX | $10–$20 | Yes (1.5x–3x common) | Expected (15–20%) |
| Lyft | $10–$19 | Yes (prime-time pricing) | Expected (15–20%) |
| Waymo (with tip factored in for Uber/Lyft) | $12–$18 | — | $0 |
| UberX (with 18% tip) | $12–$24 | — | Included |
The total cost comparison often favors Waymo when you account for tipping, which is customary for human-driven rideshare but not applicable to robotaxis. Zoox pricing has not been widely published as its commercial operations remain limited. Waymo occasionally offers promotional pricing (e.g., first rides discounted or free) to build adoption in new markets.
How to use it: Download the Waymo One app (available in service cities), enter your destination, confirm pickup, and a driverless Jaguar I-PACE will arrive. The app shows the vehicle's approach in real-time. Once inside, a screen displays your route, and you can start the ride or contact support. There is no driver to interact with. Zoox rides, where available, are requested through the Zoox app and feature the company's distinctive carriage-style vehicle with facing bench seats.
The Regulatory Landscape: Who Approves Robotaxi Operations
Autonomous vehicle regulation in the U.S. is a patchwork of federal and state authority, and understanding this structure is essential for evaluating autonomous vehicle safety claims.
Federal Level
NHTSA (National Highway Traffic Safety Administration) sets vehicle safety standards (FMVSS) and investigates crashes. However, NHTSA does not currently "approve" autonomous driving systems before deployment the way the FAA approves aircraft. Instead, it relies on post-market oversight: Standing General Orders for crash reporting, investigations into specific incidents, and the authority to issue recalls. The NHTSA ADS safety page maintains public resources on its regulatory approach.
State Level
States control where and how robotaxis can operate commercially:
- California: The California DMV issues testing permits and deployment permits. The California Public Utilities Commission (CPUC) issues commercial passenger service permits. Both must be obtained for a company like Waymo to operate commercially in San Francisco or Los Angeles.
- Arizona: Among the most permissive regulatory environments. Waymo has operated commercially in Phoenix since 2020 with relatively fewer state-level restrictions.
- Texas: Austin has emerged as a new market, with Waymo launching commercial service in late 2025 under a state framework that allows ADS-equipped vehicles without traditional safety equipment (steering wheels, mirrors) under specific conditions.
- Nevada: Las Vegas hosts Zoox operations under a state regulatory framework that has permitted driverless testing and limited commercial service.
This state-by-state approach means that autonomous vehicle safety standards, reporting requirements, and operational restrictions vary significantly depending on where you ride. A robotaxi that is legal and operational in Phoenix may face entirely different regulatory requirements—or be prohibited entirely—in a neighboring state.
Current Limitations: What Robotaxis Cannot Do
Level 4 robotaxis are genuinely driverless within their geofenced domains, but they are far from capable of handling all driving scenarios. The following limitations are current as of early 2026 and represent hard operational boundaries, not temporary inconveniences.
- Heavy precipitation: Torrential rain, heavy snow, and dense fog degrade LiDAR and camera performance. Waymo vehicles may pause operations, pull over safely, or restrict service areas during severe weather. Phoenix's dry climate is not coincidental—it's an ideal ODD for current sensor technology.
- Unmapped areas: The vehicle will not drive outside its geofence. Construction zones that significantly alter road layouts may cause the vehicle to stop and request remote assistance or reroute.
- Complex unprompted interactions: A police officer directing traffic with hand signals, a crossing guard, or a construction worker using gestures may not be interpreted correctly. The vehicle may stop and wait or request remote guidance.
- Off-road and non-standard roads: Dirt roads, parking garages with tight clearances, drive-through lanes, and private property are generally outside the ODD.
- Passenger emergencies: If a rider experiences a medical emergency, there is no driver to assist. The vehicle has an in-cabin support button connecting to a remote operator who can call 911, but cannot provide physical aid or deviate from the route to reach a hospital directly unless specifically programmed to do so.
- Accessibility limitations: Current Waymo vehicles (Jaguar I-PACE) require riders to enter a standard SUV. Wheelchair-accessible robotaxis are not yet in commercial service, though Zoox's purpose-built vehicle with its flat floor and wide doors was designed with accessibility in mind.
- Limited cargo and pets: Policies vary, but generally small items and pets in carriers are allowed. Large cargo, uncontained pets, or situations requiring driver assistance (loading a bicycle, for example) are not supported.
Autonomous Vehicle Safety: Frequently Asked Questions
Where can I use a robotaxi service right now?
As of early 2026, Waymo One operates commercial driverless rides in Phoenix (metro-wide), San Francisco, Los Angeles (expanding), and Austin (launched late 2025). Zoox offers limited commercial rides on a defined corridor in Las Vegas. Outside these cities, no Level 4 robotaxi service is available to the public. Tesla FSD, GM Super Cruise, and Ford BlueCruise are Level 2 systems that require a licensed, attentive driver—they are not robotaxis.
How much does a robotaxi cost vs. Uber or Lyft?
Waymo fares typically range from $12–$18 for a 5-mile urban trip, comparable to UberX or Lyft before tip. Because there is no driver to tip, the total out-of-pocket cost is often lower than rideshare when customary 15–20% tips are included. Robotaxis do not apply surge pricing, though wait times may increase during peak demand.
How safe are robotaxis based on real data?
Waymo's published data, based on over 7.1 million driverless miles, shows a 76% reduction in injury-causing crashes and a 73% reduction in property-damage crashes compared to human benchmarks in the same cities. NHTSA's Standing General Order crash data shows that most Level 4 robotaxi incidents are low-severity, often involving the robotaxi being struck by human-driven vehicles. No pedestrian fatalities have been recorded in Waymo's commercial driverless operations. However, the total autonomous mileage remains a small fraction of human-driven mileage, and long-term fatality-rate comparisons require significantly more data.
What are the current limitations of robotaxis?
Robotaxis cannot operate in heavy snow, dense fog, or torrential rain. They are confined to pre-mapped geofenced areas and cannot handle off-road driving, complex human gesture-based traffic control, or passenger medical emergencies. Wheelchair-accessible vehicles are not yet in commercial robotaxi service. Service is limited to specific cities—most of the U.S. has no robotaxi access.
Is this truly self-driving or just driver-assist?
Waymo and Zoox robotaxis are SAE Level 4—genuinely driverless within their defined operational domains. No human is present to drive. Consumer systems like Tesla FSD Supervised, GM Super Cruise, and Ford BlueCruise are SAE Level 2—partial automation requiring continuous human supervision. These are fundamentally different categories with different safety profiles, liability frameworks, and regulatory oversight. Conflating the two is the single most common source of public confusion about autonomous vehicle safety.
Sources: NHTSA Automated Vehicles Safety, Waymo Safety Report, California DMV Autonomous Vehicle Disengagement Reports.
