The current status of every US robotaxi operator — who's live, in which cities, at what scale — plus which production EVs offer hands-off or eyes-off driving. Verified monthly against operator press, SEC filings, and state DMV records.
Every operator running paid or near-paid public robotaxi service in the US as of August 2026. Status tracks how far a service has got toward unsupervised public operation — not how big it is. Active = driverless, no safety operator aboard, and open to the general public in at least one market. Limited = open to the public but still supervised, or driverless for only a subset of vehicles or a restricted group. Preparing = permitted or testing, not yet carrying public riders. A large supervised fleet is still a supervised fleet, so fleet size and city count are reported separately and never set this label.
Waymo
Alphabet · L4 robotaxi (no human in vehicle)
Active
The clear US robotaxi leader — public rides in 11 US metros and ~500K paid rides per week. Freeway rides are returning after a two-month pause: Waymo began restoring them on July 29, 2026, starting with Phoenix.
Cities
Phoenix, AZ · San Francisco Bay Area, CA · Los Angeles, CA · Atlanta, GA · Austin, TX · Dallas, TX · Houston, TX · San Antonio, TX · Orlando, FL · Miami, FL · Nashville, TN
How to book
Waymo One app (Phoenix, San Francisco, Los Angeles, Miami, Orlando, Nashville) · Uber app (Austin, Atlanta — exclusive in those metros) · Lyft app (Nashville)
Vehicle
Jaguar I-PACE (primary in-service vehicle); 6th-gen Zeekr-built robotaxi now carrying public riders as the 'Ojai' (since May 28, 2026); Hyundai IONIQ 5 still ramping toward fleet service
Scale
Fleet: 3,500
Weekly rides: 500,000
Weekly autonomous mi: 4,000,000
Expansion: on July 8-9, 2026 Waymo began employee-first, fully-driverless rides in Las Vegas, NV — the first of the four new US markets it announced on July 8, 2026; Denver, San Diego, and Tampa are still to follow, starting employee-first in the coming weeks with public service targeted by end of 2026 (plus London later in 2026). Waymo reaffirmed its 20+ cities and 1M weekly trips goals; these new markets are employee-first, so they are not yet counted as public live cities here. Scale is approximate — July 2026 reports put the in-service fleet at roughly 3,500-4,000 vehicles, and Waymo has now passed 20M+ lifetime paid trips. Operational status: Waymo paused public freeway rides in late May 2026 (surface-street service was unaffected throughout) and began restoring them on July 29, 2026 after software updates addressing freeway construction zones — enhanced scene recognition plus routing changes, remedying recall 26E035. The restoration is phased rather than instant: Phoenix went first on July 29, which also restored freeway routing to Sky Harbor, and Waymo said Los Angeles and the San Francisco Bay Area would follow within days. Cities beyond Phoenix are announced rather than independently confirmed live as of August 1, 2026. Regulatory: NHTSA's July 31, 2026 deadline for AV first-responder remediation plans has passed; no operator's submission has been made public and NHTSA has announced no outcome (see the US regulation section). Rider experience: on July 29, 2026 Waymo added Google's Gemini assistant and a redesigned adaptive three-screen interface (including a 'Calm Mode') to the Ojai — Waymo states Gemini handles cabin controls, trip information and general questions only, does not control vehicle movement or routing, and has no access to real-time driving data. Its 6th-gen Zeekr-built vehicle (the 'Ojai') began carrying public riders on May 28, 2026 in LA, Phoenix, and SF; the Hyundai IONIQ 5 is still ramping toward fleet service (Arizona production expected late 2026). Launched fully autonomous public rides in Nashville, TN on April 7, 2026 (Waymo One app + Lyft). Added a paid 'Waymo Premier' membership ($29.99/mo, invite-only) on June 11, 2026. Airport access includes Phoenix Sky Harbor and SFO. Going international: raised ~$16B (Feb 2026) with London named as its first overseas market and testing underway in Tokyo. Pauses operations during heavy rain or flooding in some metros until conditions clear.
Independent safety research: The first independent per-mile crash benchmark of a driverless fleet was published by the Insurance Institute for Highway Safety (IIHS) on July 23, 2026. Measured per vehicle mile traveled against human drivers in the same cities and years, Waymo's rider-only vehicles were involved in 68% fewer police-reportable crashes overall, 85% fewer single-vehicle crashes, and 81% fewer injury crashes. The metro-level picture is uneven and worth stating plainly: 76% lower in Phoenix, 71% lower in Los Angeles, and 35% lower in San Francisco — but 4% HIGHER in Austin, where IIHS notes the sample was relatively small and draws no firm conclusion. Scope: roughly 50 million Waymo rider-only miles (2021-2024) compared against roughly 222 billion human-driven miles across those four metros. Method: built from NHTSA Standing General Order crash reports, with redundant and non-qualifying records culled and a police-reportability standard then applied — only about 22% of the 736 automation-engaged public-road crashes were judged likely police-reportable, and of Waymo's 89 police-reportable involvements 64 occurred in fully driverless operation. Three caveats are essential to reading this honestly: (1) Waymo is the only operator that publishes the vehicle-miles-traveled denominator, and it supplied that mileage voluntarily, so no other robotaxi operator can currently be benchmarked this way — this finding says nothing about Tesla Robotaxi, Zoox, or anyone else; (2) the Austin sample is small; and (3) IIHS itself cautions that the present data-collection system 'isn't good enough to allow continuous monitoring of a large-scale expansion.' IIHS is funded by the auto-insurance industry; there is no indication that Waymo funded the research.
Tesla Robotaxi
Tesla · Geofenced robotaxi (mixed supervised + unsupervised)
Limited
Seven metros across three states — the widest US footprint after Waymo — but the driverless part is small. Tesla began removing safety monitors in Austin in January 2026 on a few vehicles "mixed in with the broader robotaxi fleet with safety monitors", and every independent count since has put the unsupervised subset in the low tens even as the map expanded. That is why this reads Limited rather than Active: reach is broad, unsupervised operation is not.
Cities
Austin, TX · Dallas, TX · Houston, TX · Miami, FL · Orlando, FL · Tampa, FL · San Francisco Bay Area, CA (safety driver — CA law)
Preparing
Phoenix, AZ — Tesla-confirmed 'Preparations Underway' (Q2 2026 disclosure; no launch date given) · Las Vegas, NV — PERMITTED but not yet carrying public riders: the Nevada Transportation Authority granted Tesla Robotaxi, LLC an Autonomous Vehicle Network Company permit on July 27, 2026, capped at 10 vehicles on the Las Vegas Strip corridor. Tesla had requested 5,000.
How to book
Tesla app — Robotaxi
Vehicle
Tesla Model Y. The purpose-built Cybercab is in production at Giga Texas (installed annual capacity above 125,000 units per Tesla's Q2 2026 manufacturing table) and began carrying employee riders on the Giga Texas campus in July 2026 after public-road engineering test drives in Q2 — but it is NOT yet in public Robotaxi service, and Tesla has disclosed no date for volume production or for Cybercabs entering the passenger fleet.
Scale
Fleet: Not disclosed by Tesla. Independent counts through mid-2026 span roughly 20 to 59 vehicles depending on what is being counted — Bloomberg put the total Texas robotaxi fleet at 59 (June 10, 2026), while trackers counting cars actively running reported far fewer. The one documentary figure is 42 vehicles authorized in Texas state filings (electrive, June 2, 2026). The SF Bay Area supervised fleet is larger (~100+ per late-2025 reporting). Florida fleet sizes undisclosed. Morgan Stanley projects ~1,500 vehicles by end-2026.
Weekly rides: not disclosed
driverlessVehicles: Tesla does not disclose it, and no independent count covering all seven metros has been published. Every count through mid-2026 puts the unsupervised subset in the low tens — small enough that the driverless share of the map is a fraction of it. We deliberately do not publish a precise figure here: the available counts come from different outlets measuring different things (total fleet vs. state-authorized vs. actively running vs. running unsupervised), they disagree with each other, and the most-cited of them predate the July 2026 Florida launches.
unsupervisedMiles: More than 380,000 cumulative unsupervised miles across six cities in two states — Tesla's own figure, given by VP of AI software Ashok Elluswamy on the Q2 2026 earnings call (July 22, 2026). For scale, Waymo books roughly 4 million rider-only miles per week, so Tesla's entire cumulative unsupervised mileage is under one day of Waymo driving.
Architecturally distinct from Waymo: Tesla deploys the same vision-only neural network used in consumer FSD inside a geofence, rather than relying on HD-mapped cities and lidar — that's the bet that scaling is faster, even if early-rollout reliability metrics lag a more mapped, sensor-rich approach. Florida rollout (July 2026): Tesla went fully driverless from day one in Miami on July 3 — its first market outside Texas and California and the first time it skipped the safety-monitor phase in a new city — then added Orlando and Tampa on July 21, both also driverless day one and hailable through the public Robotaxi app. Per trade press, the new Florida geofences are small: Orlando runs roughly 12-15 sq mi across eastern/southeastern neighborhoods (Conway, Lee Vista) bounded by the Beachline Expressway (SR-528) and the Central Florida GreeneWay (SR-417); Tampa runs roughly 4-5 sq mi over West Tampa, Tampa Heights, and Seminole Heights north of downtown, with the airport excluded (Miami's zone is ~10-20 sq mi of western Miami-Dade near the airport, West Miami, and Doral, excluding downtown and Brickell). The Texas map footprint is large but thin: state filings list 42 authorized vehicles, while independent trackers counting cars actively running reported materially fewer through mid-2026. The SF Bay Area service (running since 2025) is Tesla's largest by vehicle count but requires a safety driver — California still classifies it as a limousine offering rather than a robotaxi and has an open state action over 'Full Self-Driving' marketing. Tesla's own Q2 2026 disclosure makes that concrete: it lists the SF Bay Area status as 'Safety Driver' and cites California charter-party carrier permit #TCP0046782-A — a limousine/chauffeured-transport authorization, not an autonomous-vehicle deployment permit, which is exactly why a human sits behind the wheel there. Musk has tied aggressive scaling to FSD v15 — a model rewrite growing from ~1B to ~10B parameters — pointing to late 2026 / early 2027. On the Q2 2026 earnings call (call-sourced; this detail is not in the written deck) Tesla said the robotaxi fleet is already running early v15 builds, with roughly 40% of 7 planned improvement tracks merged. That applies to the geofenced robotaxi fleet, not to consumer cars: large-scale unsupervised deployment to customer vehicles is still late-2026-at-earliest / early-2027-at-latest. Tesla's Q2 2026 disclosure also states it expanded the unsupervised operating area in Austin during the quarter, and publishes a per-metro status table listing Austin, Dallas, Houston, Miami, Orlando and Tampa as 'Ramping Unsupervised' and the SF Bay Area as 'Safety Driver'. Next markets, now Tesla-confirmed rather than rumored: that same table lists Phoenix, AZ and Las Vegas, NV as 'Preparations Underway' — Tesla's own language — though it gives no launch date for either, so none should be inferred. Supporting groundwork already on the record: on June 5, 2026 Tesla filed for a Nevada Autonomous Vehicle Network Company permit covering Las Vegas (including Harry Reid and Henderson airports) and allowing up to 5,000 vehicles in year one. The Nevada Transportation Authority granted that application on July 27, 2026 (AVNC Permit 002, docket 26-05015) — but capped operations at ten fully autonomous vehicles inside an Authority-approved geofence limited to the Las Vegas Strip corridor, with all trips originating and terminating in Clark County; barred passenger transport on Nevada roads posted above 45 mph; required every vehicle to be marked "Robotaxi" and every passenger notified before each trip that the ride is driverless; mandated "appropriate human supervision" per SAE standards; and prohibited pickups at or within a quarter-mile of Harry Reid International Airport absent the airport operator's authorization. So Tesla received 0.2% of the fleet it asked for, and the airport access it sought was not granted. Expanding the geofence or fleet requires fresh Authority approval under NRS 482A.100; in Arizona it has filed a 'Robotaxi First Responder Interaction Plan' outlining a proposed Phoenix-metro service footprint, and trade-press reports put roughly 60 robotaxi-spec Model Ys staged in the Phoenix area. Neither metro is carrying public riders yet.
Zoox
Amazon · L4 robotaxi — purpose-built vehicle (no steering wheel or pedals)
Active
Amazon-owned, purpose-built robotaxi. Rides on the Las Vegas Strip stop being free on Monday, August 10, 2026, the date Zoox starts charging fares — its first commercial market, and the first paid service anywhere in the US from a vehicle built with no steering wheel or pedals, made possible by the NHTSA exemption granted July 30, 2026.
Cities
Las Vegas, NV · San Francisco, CA (public app waitlist — free) · Austin, TX (early access — employees, family and friends; not open to the public)
How to book
Zoox app
Vehicle
Zoox purpose-built carriage-style robotaxi (bidirectional, no driver controls)
Scale
Total autonomous mi: nearly 2 million as of March 2026 (Zoox figure; no newer count published, and riders have since grown from ~350,000 to nearly 1 million, so the true figure is higher)
Total passengers: nearly 1 million riders across Las Vegas, San Francisco, Austin and Miami since the free service launched (Zoox figure, reported Aug 5, 2026); Las Vegas alone accounts for 500,000+ rides
Zoox now charges for rides. Because its vehicle has no steering wheel or pedals, Zoox needed an NHTSA exemption from the Federal Motor Vehicle Safety Standards before it could operate commercially — NHTSA granted one on July 30, 2026, covering up to 2,500 vehicles annually for two years, waiving eight FMVSS (including manual driving controls, mirrors, windshield defrosting and light-vehicle braking systems) and carrying what NHTSA calls an 'enhanced, adaptable oversight structure.' NHTSA describes it as the first commercial exemption of its kind for a purpose-built robotaxi, which is what makes the paid launch a US first for a controls-free vehicle. On August 5, 2026 Zoox announced that paid service in Las Vegas starts Monday, August 10, 2026 — its first commercial market after roughly a year of free public rides. How the fare works, per Zoox: a base fare plus distance and time from pickup to drop-off, calculated on the best available route at the moment of booking and shown in full before the rider confirms; if the vehicle then takes a different or longer route, the rider still pays the quoted price. Destination fees apply on top for certain high-traffic locations — Harry Reid International Airport, Sphere and T-Mobile Arena. Zoox declined to publish a base fare and would only say it aims to be competitive with the 'comfort' tier of ride-hailing (the tier above UberX); no verified dollar figure or percentage exists yet, so none is stated here. Airport rides are still test-only, so the airport destination fee applies once public airport service opens. Paid service is Las Vegas only for now: Zoox needs state and local approvals to charge fares anywhere else, and California specifically still requires driverless deployment permits from the CPUC and the DMV, so the federal exemption alone does not open that market. Las Vegas service area more than doubled in March 2026 (Convention Center + Strip hotels). San Francisco is open to the general public, but gated and free. Anyone can download the Zoox app and join the waitlist; Zoox invites riders off it into its "Zoox Explorers" early-rider program, which replaced employee-only access on November 18, 2025. Fares there are not merely unannounced — they are not permitted: the CPUC lists Zoox (TCP 38380) with Drivered Pilot and Driverless Pilot permits only and no Deployment permit, and CPUC rules bar monetary compensation for rides under the pilot programs; the California DMV likewise lists Zoox with a driverless testing permit (San Francisco and Foster City) and no deployment permit. The Las Vegas Review-Journal reported on August 5, 2026 that Zoox has no plan to begin paid service in San Francisco. Its SF service area quadrupled in March 2026 (SoMa, Mission and Design District at launch). Austin is carrying riders but is likewise limited to Zoox employees and their family and friends, with the 'Zoox Explorers' early-rider program planned later in the year — it is not a public market. Miami is also announced as a launch market with a small early-access zone. Unveiled a redesigned production robotaxi on June 24-25, 2026 ahead of series production; its Hayward, CA factory is scaling toward up to 100 vehicles/week. Booking is through the Zoox app; the previously stated summer 2026 arrival on the Uber app has slipped, with Zoox now saying rides become bookable on Uber later this year.
Motional
Hyundai (sole owner after Aptiv exit) · L4 robotaxi — live supervised pilot, driverless targeted EOY 2026
Limited
Hyundai-owned, rebooted in early 2026. Hailable on the Uber app across five Las Vegas zones — but with a safety monitor onboard in every vehicle, and driverless service still only targeted for the end of 2026. Reclassified from Active to Limited on 2026-08-21: a fully supervised pilot in one city is by definition less far along than services already running unsupervised, and labelling it Active while Tesla read Limited was inconsistent.
Cities
Las Vegas, NV (supervised pilot; driverless targeted EOY 2026)
How to book
Uber app (Las Vegas)
Vehicle
Hyundai IONIQ 5 robotaxi
Scale
Phase: supervised public pilot — safety monitor still onboard
As of March 13, 2026, Motional's IONIQ 5 is bookable on the Uber app across five Las Vegas zones (Resorts World, Encore, Westgate, Town Square, Downtown) with a safety monitor onboard — a live supervised public pilot, the 'final proving ground' before removing the monitor for the targeted end-of-2026 driverless launch. Aptiv exited the joint venture in 2025; Hyundai is the sole owner and added ~$1B in the reboot (which included deep layoffs and an AI-first perception stack).
May Mobility
Independent (Toyota / BMW i Ventures / others as investors; ECARX partnership) · Shared-shuttle AV service
Active
The other US AV operator — lower-speed transit/paratransit, and now running a fully driverless (driver-out) rider service in the Atlanta metro.
Cities
Peachtree Corners, GA (driverless) · Atlanta, GA (via Lyft) · Grand Rapids, MN · Eden Prairie, MN
How to book
Local transit-style booking via partner agencies; Lyft integration in some markets
Vehicle
Toyota Sienna AV minivans (transitioning toward lower-cost platforms with ECARX)
Scale
Focus: Lower-speed shared shuttles, often as on-demand transit or paratransit
footprint: Four active US deployments plus Nagoya, Japan, per May Mobility's own locations page (checked August 21, 2026). Earlier Sun City, AZ and Detroit, MI deployments are no longer listed among active deployments.
Operates driver-out (no safety operator aboard) in Peachtree Corners, Georgia — May Mobility's own locations page labels that service "(Driverless)" — which is what makes this row Active. It launched there in February 2025 as the company's first commercial driver-out service, with Curiosity Lab, T-Mobile and the City of Peachtree Corners; the service is free and runs weekdays 9:30am-3pm. Its other current US deployments (Atlanta via Lyft, Grand Rapids MN, Eden Prairie MN) carry an onboard operator. In May 2026 it signed a framework deal with ECARX worth roughly $750M to supply autonomy-enabled vehicles. Footprint note: May Mobility's earlier Sun City, AZ and Detroit, MI deployments no longer appear among its active deployments as of August 21, 2026 — the company's US footprint has consolidated rather than grown, which is worth weighing against the ECARX headline number.
Avride
Independent (spun out of Yandex) · L4 robotaxi — live via Uber (safety operator onboard)
Limited
Uber's Dallas robotaxi partner, and the busiest US service that is still not driverless: more than 100,000 autonomous rides on the Uber app as of August 2026, every one of them with a safety operator behind the wheel.
Cities
Dallas, TX
How to book
Uber app (Dallas)
Vehicle
Hyundai IONIQ 5
Scale
Phase: early commercial — safety operator still onboard, driver-out planned
rides: More than 100,000 autonomous rides on Uber in Dallas, announced August 14, 2026.
Fleet: About 200 vehicles. Avride states 'over 200'; NHTSA independently estimates the affected population at 200 in its open investigation — two sources agreeing on the same quantity.
miles: Avride reports more than 1.3 million cumulative autonomous miles, roughly 1 million of them in 2026. Company-stated; not independently verified.
Began commercial service on December 3, 2025 in a roughly 9-square-mile zone (Downtown, Uptown, Turtle Creek, Deep Ellum) using a Hyundai IONIQ 5 fleet dispatched through Uber. Rides match normal UberX/Comfort pricing — no premium. Avride announced its 100,000th autonomous ride on Uber on August 14, 2026, and reports a fleet of over 200 vehicles and 1.3M+ cumulative autonomous miles. Driver-out remains planned rather than delivered: as of June 2026 the Dallas vehicles still carried a human safety operator behind the wheel, which is why this row reads Limited rather than Active.
REGULATORY: NHTSA's Office of Defects Investigation opened Preliminary Evaluation PE26003 into the Avride ADS on May 6, 2026, prompted by SGO 2021-01 reports and covering an estimated population of 200 vehicles. The stated subject is ADS "conflict avoidance, driving behavior competence and assertiveness" — specifically lane changes, responding to vehicles in or entering the lane ahead, and responding to stationary objects. A Preliminary Evaluation is an inquiry, not a finding of defect and not a recall. We note it because an open federal investigation into the driving stack bears directly on the driver-out timeline; NHTSA's resume also places Avride ADS operation in Austin as well as Dallas, though Dallas remains its only public Uber market.
Nuro
Nuro (partnership with Uber + Lucid) · L4 robotaxi — public-road testing (not yet carrying public riders)
Preparing
A premium robotaxi effort from Nuro, Uber and Lucid. Now carrying select Uber employees in the SF Bay Area with a safety operator aboard, with public rides targeted for late 2026 and Houston named as the second market for 2027.
Cities
San Francisco Bay Area, CA (testing) · Houston, TX (testing)
How to book
Planned via the Uber app
Vehicle
Lucid-based premium robotaxi running the Nuro Driver autonomy stack
Fleet: About 100 Lucid Gravity SUVs in the engineering fleet across California and Texas (company-stated).
Nuro pivoted from autonomous goods delivery to licensing its 'Nuro Driver' stack; this Uber + Lucid partnership puts it into premium passenger robotaxi. Autonomous on-road testing began December 2025, and Bay Area public-road testing was announced April 13, 2026. As of August 2026 the service is still pre-public: select Uber employees can hail a Lucid Gravity robotaxi through the Uber app, with a human safety operator behind the wheel. Nuro reports roughly 100 Lucid Gravity SUVs in its engineering fleet, testing 24/7 in both the SF Bay Area and Houston. Public launch is targeted for the SF Bay Area in late 2026; Houston was named the second market in June 2026 with service targeted for mid-2027, supported by a 50,000 sq ft Uber depot due to be operational by early 2027. No public riders yet, which is why this reads Preparing.
Mobileye
Mobileye (majority-owned by Intel) · L4 robotaxi — announced, not yet operating a service
Preparing
The long-time supplier of driver-assistance and autonomy tech to other automakers announced on June 16, 2026 that it will run its own robotaxi service, targeting a 2027 US launch. No city named, no vehicles carrying riders.
Cities
None yet
Vehicle
Not specified for the US service; Mobileye supplies its own Mobileye Drive autonomy stack
Scale
Phase: pre-launch — plans an initial ~100-vehicle fleet in one unnamed US city in 2027, scaling toward ~17,000 vehicles within five years of launch
Listed here for completeness because Mobileye's announcement puts it on both sides of the AV business — continuing to sell its stack to automakers and fleets while operating a competing service of its own. What is actually confirmed is limited: a vertically integrated robotaxi business, an initial fleet of roughly 100 vehicles in one major US city in 2027, and a stated ambition of about 17,000 vehicles within five years of launch. Mobileye has not named the launch city and says further commercial and operational detail will come at a US Capital Markets Day before the end of 2026. Nothing here is a live service — treat it as an announced program, not a competitor you can ride today.
No longer operating
Cruise (shut down)Shut down
GM wound down Cruise robotaxi operations in December 2024. The team and tech were absorbed into GM's personal-vehicle ADAS work (Super Cruise / next-gen).
GM cited the cost and time to scale a competitive robotaxi business as the reason for exit. Cruise is no longer operating a public robotaxi service. GM has NOT restarted a robotaxi business, but it did revive an autonomous-driving program for PERSONAL vehicles (renewed push in 2025; ex-Tesla autonomy leadership hired late 2025), and is now targeting an eyes-off (Level 3) system on the 2028 Cadillac Escalade IQ — that's a consumer-car effort, not a robotaxi relaunch. Included here so users searching the Cruise brand reach a current, sourced answer. Re-checked August 21, 2026: GM has still not restarted a robotaxi business and the Cruise service remains shut down. One nuance worth tracking — in June 2026 GM said the eyes-off system it is building for personal vehicles could eventually be capable of running as a robotaxi, so the capability may return through the personal-vehicle program rather than through a revived Cruise.
Where can I ride a robotaxi today?
US cities with public or early-access robotaxi service as of August 2026.
Waymo (public) · Tesla Robotaxi (driverless, since July 21, 2026)
Tampa, FL
Tesla Robotaxi (driverless, since July 21, 2026) · Waymo (announced July 8, 2026 — employee-first, not yet public)
Miami, FL
Waymo (public) · Tesla Robotaxi (driverless, since July 3, 2026)
Nashville, TN
Waymo (Waymo One app + Lyft)
Las Vegas, NV
Zoox (public on Strip; paid from Aug 10, 2026) · Motional (supervised pilot on Uber) · Tesla Robotaxi (preparations underway per Tesla's Q2 2026 disclosure — not yet carrying riders)
Production-vehicle autonomy — what each system actually does
Hands-free, eyes-off, and "supervised" features by automaker. No production passenger vehicle in 2026 is L4 — only robotaxis (above) are. To weigh these cars on range, price, and charging, use our EV comparison tool.
Read this section by operational scope, not by SAE level. Tesla FSD is L2 because Tesla keeps the driver legally responsible — but in functional scope it's the only system that operates on any road, including city streets, intersections, traffic lights, and unprotected turns. The hands-off L2+ systems (Super Cruise, BlueCruise, ProPILOT 2.0, Rivian Universal Hands-Free, Subaru Highway Hands-Free Assist) only engage on mapped divided highways and will not handle intersections or signals. Mercedes Drive Pilot is the only L3 (eyes-off) system certified for US roads, but only on specific highways at ≤40 mph.
Operational scope at a glance
System
Highways
City streets
Intersections
Lights & signs
Hands-off
Tesla FSD (Supervised)
Yes
Yes
Yes
Yes
No (supervised)
Mercedes Drive Pilot (L3)
Mapped highways only, ≤40 mph
No
No
No
Yes (eyes-off)
GM Super Cruise
Mapped highways only
No
No
No
Yes
Ford BlueCruise
Mapped Blue Zones only
No
No
No
Yes
Nissan ProPILOT 2.0
Mapped highways only
No
No
No
Yes
Rivian Universal Hands-Free
Yes
Lane-keep only
No
No
Yes
"Yes" = the system engages and handles that environment under normal conditions. "No" = the system either disengages or never engages there. Tesla FSD is the only production system that operates across all five environments — but it requires the driver to remain attentive and hands-on. Every other system on this list cedes those city-driving environments back to the human driver.
Tesla FSD (Supervised) / Autopilot
Tesla
L2
Level 2 (supervised) — by far the widest operational scope of any production system
Where it works: The only production system that operates on essentially ANY road. Highways + city streets + intersections + traffic lights + stop signs + unprotected turns + roundabouts + parking. Every other hands-off system on this page is restricted to a mapped highway network and explicitly will not handle intersections or traffic signals. FSD currently has 10+ billion cumulative miles driven, ~3.8 billion of those on city streets.
Vehicles: Model 3 · Model Y · Model S · Model X · Cybertruck
Safety record: Tesla publishes a quarterly Vehicle Safety Report covering the supervised fleet. Direct safety comparison to mapped-highway-only systems isn't apples-to-apples — FSD's mileage includes city driving (statistically far more dangerous per mile than divided highway). NHTSA has an active FSD investigation (escalated to an Engineering Analysis — the recall-track stage — covering ~3.2 million vehicles in 2026), and California has an open state action over 'Full Self-Driving' branding.
Availability: Available as Autopilot (standard on every Tesla), Enhanced Autopilot (one-time purchase), and FSD (Supervised) — either $99/month subscription or $8,000 one-time. 1.48 million active FSD subscriptions as of Q2 2026 — up 56% year-over-year and about 16% sequentially from Q1 2026. Tesla's disclosed metric includes both upfront purchases and monthly subscriptions and excludes free trials, so it counts paid FSD customers rather than monthly subscribers alone. Continuously updated OTA — 'v14 Lite' for older HW3/AI3 cars exited early access and reached public release on build 2026.20.6.10 on July 21, 2026, the first FSD update for HW3 owners in roughly 14 months. A unified push on build 2026.20.6.11 followed around July 23, 2026, delivering v14.3.6 to HW4/AI4 vehicles and v14 Lite to HW3/AI3 in a single release; its release notes state Tesla unified the model across Actually Smart Summon, FSD, and Robotaxi. Provenance note: these version and build numbers come from Tesla-watcher trade press quoting the release notes, not from a Tesla-published changelog.
Tesla classifies FSD as Level 2 because Tesla — not the system — accepts no legal responsibility while engaged. That is a regulatory classification, not a capability ranking: by functional scope, FSD does things no other production system attempts. It also has, by an order of magnitude, the largest real-world dataset of any consumer autonomy system. The Robotaxi product (in the operators section above) runs an unsupervised variant of the same stack inside a geofence.
Mercedes-Benz Drive Pilot (US rollout paused)
Mercedes-Benz
L3 (US rollout paused)
Level 3 — Eyes-off, hands-off (within operational design domain) — US availability paused as of early 2026
Where it works: Was approved by the California DMV (Jun 2023) and Nevada (Jan 2023). Operated on certain limited-access highways, in daylight, at speeds up to 40 mph, behind a lead vehicle, in clear weather. Not being offered on the 2026 S-Class.
Vehicles: EQS Sedan · S-Class
Availability: Paused/discontinued in the US around January 2026. Mercedes is not offering the eyes-off L3 Drive Pilot on the 2026 S-Class, shifting instead to a Level 2++ 'MB.Drive Assist Pro' hands-on system that covers a wider range of city and highway conditions. The drivers were high cost and limited customer uptake. (Often miscast as a Luminar story: Mercedes had already terminated its Luminar lidar supply deal in late 2024, and Luminar's December 2025 bankruptcy was driven mainly by a separate collapsed Volvo deal — so the pause followed it chronologically but wasn't caused by it.) No SAE Level 3 system is currently being offered to US consumers.
Drive Pilot was the first and only L3 system certified for US public roads (CA + NV) before the pause. With its US rollout halted, the US has no production L3 system currently on offer: BMW canceled its Personal Pilot L3 program globally as part of the 2026 7 Series facelift after low uptake; Honda Sensing Elite L3 remains Japan-only.
GM Super Cruise
GM (Cadillac, Chevrolet, GMC)
L2+
Level 2+ — Hands-off, eyes-on (mapped highways only — does NOT handle intersections, lights, or city streets)
Where it works: Hands-free ONLY on ~750,000 miles of mapped, divided highways across the US and Canada. Will not engage on city streets and does not handle intersections, traffic lights, or stop signs. Driver-facing IR camera enforces eyes-on requirement.
Vehicles: Cadillac Lyriq · Cadillac Vistiq · Cadillac Escalade IQ · Cadillac CT5 · Chevrolet Silverado EV · Chevrolet Equinox EV · Chevrolet Blazer EV · GMC Sierra EV · GMC Hummer EV
Safety record: GM reports zero crashes and zero fatalities across approximately 160 million Super Cruise miles. Important caveat: those miles are exclusively on mapped, divided highways — statistically the safest driving environment per mile. This isn't an apples-to-apples comparison with systems (like Tesla FSD) whose mileage includes city streets, intersections, and traffic lights.
Availability: Widely available across GM EV lineup (and many ICE models).
Conservative by design — a narrow operational scope is part of why the safety record is clean. GM's pivot away from Cruise robotaxi (December 2024) reinvested in personal-vehicle autonomy: Super Cruise passed 1 billion cumulative hands-free miles in April 2026, and GM is now testing an eyes-off (Level 3) system (200+ vehicles in CA/MI) targeted for the 2028 Cadillac Escalade IQ — highway-only, not yet in production.
Ford BlueCruise
Ford / Lincoln
L2+
Level 2+ — Hands-off, eyes-on (mapped Blue Zones — does NOT handle intersections, lights, or city streets)
Where it works: Hands-free only on Ford's pre-mapped 'Hands-Free Blue Zones' — divided highways across the US and Canada, mileage expanding via OTA. Will not engage on city streets and does not handle intersections, traffic signals, or stop signs. Outside the mapped network, BlueCruise downgrades to hands-on Adaptive Cruise + lane centering.
Vehicles: Mustang Mach-E · F-150 Lightning · F-150 (ICE) · Explorer · Expedition · Lincoln Navigator
Safety record: Public NHTSA incident reporting lists BlueCruise involved in 11 crashes with 3 fatalities since launch. Same caveat as Super Cruise: mileage is on Blue Zone highways only, so cross-system safety comparisons (especially to systems that drive in cities) aren't apples-to-apples.
Availability: Available across Ford EV and most non-EV trucks/SUVs.
Strict eyes-on enforcement via IR-based driver-monitoring camera. Scope is highway lane-keeping with automated lane changes — not a self-driving system in the full sense.
Nissan ProPILOT Assist 2.0 / 2.1
Nissan / Infiniti
L2+
Level 2+ — Hands-off on mapped highways only (no intersections, no city streets)
Where it works: Hands-free on pre-mapped divided highways in the US. Includes automated lane changes and overtaking assist. Does not engage on city streets and does not handle intersections or traffic lights.
Availability: Available as an option on higher trims; less broadly available than Super Cruise or BlueCruise.
The third pillar of mainstream hands-free-on-highway in the US, alongside GM Super Cruise and Ford BlueCruise. Same operational-scope ceiling: divided highway lane-keeping, no city driving.
Rivian Autonomy+ / Universal Hands-Free
Rivian
L2+
Level 2+ — Hands-off, eyes-on (lane-keeping only; does NOT handle intersections or traffic signals)
Where it works: Started as Enhanced Highway Assist on ~135,000 miles of approved freeways. The 2025.46 OTA expanded 'Universal Hands-Free' to 3.5M+ miles of roads including rural and some city streets — but the system does NOT stop or slow for traffic lights or stop signs. Functionally it is hands-off lane-keeping with adaptive cruise on a broader road set, not a city-driving system. The driver must take every intersection, light, and stop sign.
Vehicles: R1S (Gen 2) · R1T (Gen 2)
Availability: Subscription: $49.99/mo or $2,500 one-time. 60-day trial on new R1S/R1T deliveries.
Major scope expansion announced at Rivian's 'Autonomy Day' in 2026. Positioned as a Tesla FSD competitor in marketing, but the operational scope is meaningfully narrower — Rivian's system explicitly does not handle the intersections, signals, and unprotected turns that FSD attempts. That gap is set to narrow: 'Universal Hands-Free 2.0,' slated for Q3 2026, is announced to add stop-light/stop-sign response, surface-street automated lane change, and on-ramp-to-off-ramp routing (a step toward L3) — but it has not shipped as of late June 2026, so today the system still does not stop for lights or stop signs.
Subaru Highway Hands-Free Assist
Subaru
L2+
Level 2+ — Hands-off, eyes-on (highways only — does NOT handle intersections, lights, or city streets)
Where it works: Hands-free, eyes-on on most controlled-access US highways at speeds up to ~85 mph, with a driver-monitoring camera enforcing attention. Like every other L2+ system here, it does not engage on city streets and does not handle intersections, traffic lights, or stop signs.
Vehicles: Outback Touring · Outback Touring XT
Availability: Subaru's first hands-free system, announced March 2026 on the 2026 Outback Touring / Touring XT — notably offered with no-cost OTA/dealer activation for eligible existing owners.
A genuine new entrant to the US hands-free-on-highway club (alongside GM Super Cruise, Ford BlueCruise, Nissan ProPILOT 2.0/2.1, and Rivian). Same operational ceiling: highway lane-keeping, no city driving.
BMW Personal Pilot L3 (discontinued)
BMW
L3 (discontinued)
Level 3 — never offered in US; discontinued globally
Where it works: Germany only — operated up to 60 km/h (37 mph) on divided highways in daylight, no construction zones or tolls.
Vehicles: 7 Series (Germany only, 2023–2026)
Availability: Discontinued. The 2027 7 Series facelift drops L3 in favor of a Level 2 'Symbiotic Drive' hands-off / eyes-on system that debuted in the new iX3.
BMW publicly cited insufficient customer demand to justify continued L3 development. Listed here for query coverage — readers searching 'BMW Personal Pilot US' deserve a clean answer.
SAE autonomy levels — the quick reference
The SAE J3016 framework is the global standard for talking about vehicle automation. The single most useful distinction: at L2 and below, the driver is legally responsible — at L3 and above, the system is, while engaged.
An important nuance often missed: the SAE level reflects who accepts legal responsibility, not how capable the system is. A more-capable system can sit at a lower SAE level if the manufacturer chooses to keep liability with the driver. Tesla FSD is the clearest example — classified L2 because Tesla doesn't accept liability while engaged, yet it handles intersections, traffic lights, stop signs, and city streets that every L2+ hands-off competitor explicitly refuses to operate on.
L0
No automation
Driver does everything. Adaptive cruise control alone is L0/L1 depending on integration.
System controls either steering OR speed, never both.
Examples: Adaptive cruise control without lane-keeping. Lane-keeping without ACC.
L2
Partial automation
System controls steering AND speed. Driver must monitor and be ready to take over at any time, hands on or off the wheel depending on the specific system. The SAE level reflects who is legally responsible — it is NOT a ranking of how capable the system is. Tesla FSD is L2 because Tesla doesn't accept liability while engaged, but functionally it handles intersections and city streets that mapped-highway L2+ systems explicitly will not.
Examples: Tesla Autopilot, Tesla FSD (Supervised). Most consumer ADAS sits here.
L2+
Hands-off, eyes-on (industry term, not formal SAE)
Same legal status as L2 — driver is responsible — but the system permits hands-off operation in a defined road network, with eye-tracking enforcing attention. In practice this means lane-keeping on mapped divided highways. None of the L2+ systems below handle intersections, traffic lights, stop signs, or city streets.
Examples: GM Super Cruise, Ford BlueCruise, Nissan ProPILOT 2.0/2.1, Rivian Universal Hands-Free.
L3
Conditional automation
Eyes OFF allowed within a strict operational design domain (specific roads, speeds, weather). The SYSTEM is responsible while engaged; the driver must take over when prompted.
Examples: Mercedes-Benz Drive Pilot (was the only L3 certified for US public roads, CA + NV) — but its US rollout was paused in early 2026, so no L3 system is currently offered to US consumers.
L4
High automation
No human driver required within the operational design domain. The vehicle handles all edge cases or pulls over safely.
Examples: Waymo, Zoox, Motional (target), Tesla Robotaxi (within geofence).
L5
Full automation
No operational design domain — any road, any condition, no human input ever needed.
Examples: No production system exists at L5 as of 2026.
International operators (not yet in the US)
Listed for query coverage — readers searching these operators deserve a current, sourced answer about US availability.
Baidu Apollo Go(China (+ Middle East, Asia, Europe pilots)) — The world's largest robotaxi operation — peaked above ~350,000 weekly driverless rides in March 2026 across 27+ cities, now just ahead of Waymo globally. Went international in 2026 (Dubai driverless via Uber, Abu Dhabi, Seoul, a Switzerland L4 permit). No US service.
WeRide(China / UAE / Saudi / Europe) — Launched fully driverless, fare-charging robotaxi service in Dubai via Uber on March 31, 2026 (no onboard operator); also Abu Dhabi and Riyadh, with a 1,200+ vehicle Middle East commitment by ~2027. US presence is a California test permit only (no public service).
Pony.ai(China / Europe / Middle East) — Robotaxi service in multiple Chinese cities; launched Europe's first commercial robotaxi service in Zagreb, Croatia (early 2026, with Verne + Uber) and is expanding toward Dubai/Doha. Ranked among the top global operators. No US public service.
AutoX(China) — One of the first operators anywhere to run fully driverless robotaxis, opening a no-safety-driver service in Shenzhen in 2021, and still operating in Shenzhen and Shanghai. AutoX took a California driverless testing permit in 2020 but never launched a US service, and it no longer appears on any of the California DMV's current permit lists — drivered testing, driverless testing, or deployment. No US service.
Wayve(UK) — AI-first AV stack. With Uber + Nissan, will begin preparing a Tokyo robotaxi pilot (Nissan LEAF) by late 2026; London remains pre-commercial. Has had US ADAS testing in the SF Bay Area since 2024 (no US robotaxi service) and announced a 2026 Stellantis + Uber L4 partnership.
US autonomy regulation — what changed recently
The rules move almost as fast as the operators. The federal and state actions below directly shape where and how robotaxis can scale.
DOT/NHTSA AV Framework — national performance standards + Zoox exemptionAnnounced; FMVSS updates pending, interim guidance out for commentJuly 30, 2026 (Federal Register notice published July 31, 2026)
Transportation Secretary Sean Duffy directed NHTSA to accelerate development of the first national AV performance standards and to loosen the exemption path for vehicles built without manual controls. Three concrete pieces: (1) NHTSA granted Zoox a temporary commercial exemption — up to 2,500 vehicles annually for two years — the first of its kind for a purpose-built robotaxi; (2) NHTSA stated its intention to update FMVSS No. 101 (controls and displays), 108 (lighting), 111 (mirrors and rearview display), 126 (electronic stability control), and 201/208 (sun visors and warning labels), with the stated goal that once complete, individual vehicle exemptions may no longer be necessary; (3) a Federal Register notice published July 31, 2026 requests public comment on interim guidance for commercial deployment exemptions. This supersedes the withdrawn AV STEP approach below as the operative federal direction. The comment deadline is not stated here because it could not be verified against the notice text directly.
AV Emergency Response Coordination ActIntroduced — not lawJuly 28, 2026
Rep. Kevin Mullin (D, CA-15) introduced a bill that would set the first binding federal requirements for how autonomous vehicles behave around emergencies. It would require AV operators to publish clear emergency protocols for first responders, staff a 24-hour hotline giving public officials direct access to the company, direct NHTSA to develop minimum national emergency-response standards, and create a process letting public officials geofence AVs — restrict where they operate — during emergencies or ongoing safety hazards. Mullin announced it at a San Francisco press conference alongside the mayor, the fire and police chiefs, SFMTA, and Advocates for Highway and Auto Safety. Like the SELF DRIVE Act below, it is introduced legislation only — track as pending, not law.
NHTSA directive — AV first-responder remediation plansDeadline passed July 31, 2026 — submissions not public, no NHTSA outcome announcedJuly 8, 2026 (deadline July 31, 2026)
NHTSA Administrator Jonathan Morrison publicly directed autonomous-vehicle developers to submit remediation plans by the end of July 2026 for robotaxis interfering with active emergency scenes — entering emergency areas, blocking ambulances or fire trucks, or failing to respond to flashing lights, flares, smoke, and cones. Morrison characterized this as a 'functional insufficiency' rather than a rare edge case, with enforcement on the table, and said NHTSA would meet with developers before month-end. It is widely read as aimed at the largest US operator, Waymo. The directive followed operational failures including a July 4, 2026 event in which multiple Waymo vehicles were stranded in a congested San Francisco area (near the Presidio) and required assistance to clear. NHTSA also signaled it will seek public and industry comment toward defining objective 'behavioral competencies' and test standards for how AVs perform around emergency scenes. Status update: the July 31, 2026 deadline has now passed. No operator's submission has been made public, and NHTSA has announced no outcome or enforcement action — so nothing is currently known about whether or how developers responded. Absence of a public filing is not evidence of non-compliance.
NHTSA proposed amending Federal Motor Vehicle Safety Standard No. 135 to drop manual brake-control (pedal) requirements for vehicles designed to never have a human driver — i.e. purpose-built robotaxis like Zoox. Braking-performance and stopping-distance standards are preserved; nothing changes for vehicles with manual controls. Comment period runs to July 27, 2026. A meaningful step toward letting no-steering-wheel/no-pedal robotaxis operate at scale.
NHTSA withdraws the AV STEP evaluation frameworkWithdrawnJune 26, 2026
NHTSA withdrew the proposed 'AV STEP' program — the Biden-era voluntary national framework for evaluating automated-driving systems. The federal approach is shifting toward targeted FMVSS modernization (above) rather than a new voluntary evaluation regime.
SELF DRIVE Act of 2026 (H.R. 7390)Pending — not lawIntroduced February 2026
The first dedicated federal autonomous-vehicle bill of this Congress. Introduced February 5, 2026 and advanced out of subcommittee February 10 on a narrow 12–11 vote. Aims to set a national framework (and FMVSS exemptions) for AVs, but is not enacted — track as pending.
California DMV — new AV rules + enforcementFinalizedApril 28, 2026
California finalized rules that lift the ban on autonomous vehicles over 10,001 lbs (allowing driverless freight trucks), permit medium-duty AV transit, and add enforcement teeth — a roadside 'Notice of AV Noncompliance' to manufacturers begins July 1, 2026 (72-hour response), plus geofencing directives and a 30-second first-responder response requirement.
Texas SB 2807 — AV permit requirementIn effectEffective Sept 1, 2025; commercial authorization required from May 28, 2026
Texas's autonomous-vehicle law took effect September 1, 2025; the requirement for a state (TxDMV) authorization before commercial driverless operation became enforceable May 28, 2026. Relevant because Texas hosts Waymo, Tesla Robotaxi, and Avride.
Waymo serves 11 US metros open to public riders: Phoenix, San Francisco, Los Angeles, Atlanta, Austin, Dallas, Houston, San Antonio, Orlando, Miami, and Nashville (freeway rides, paused since late May 2026, began returning July 29, 2026 starting with Phoenix, with Los Angeles and the SF Bay Area announced to follow; surface-street service ran normally throughout). Zoox runs a paid service on the Las Vegas Strip from August 10, 2026 (NHTSA cleared it to charge on July 30), plus a free service in San Francisco that anyone can join via its app waitlist. Motional is a supervised pilot on the Uber app in Las Vegas. Tesla Robotaxi operates in seven US metros — the Austin metro, Dallas, and Houston (Texas), plus Miami, Orlando, and Tampa (Florida) — though only a subset of vehicles in those markets runs without a safety monitor, with supervised cars mixed into the same fleet — and the San Francisco Bay Area with a safety driver (California law). Avride runs on the Uber app in Dallas with a safety operator behind the wheel.
Is Tesla FSD self-driving?
Functionally, Tesla FSD (Supervised) handles more driving than any other production system: city streets, intersections, traffic lights, stop signs, unprotected turns, parking, and highways — essentially any road. By SAE classification it is Level 2, but that classification reflects legal responsibility (Tesla doesn't accept liability while engaged), not capability. The hands-off L2+ competitors (GM Super Cruise, Ford BlueCruise, Nissan ProPILOT 2.0, Rivian Universal Hands-Free) are restricted to mapped divided highways and will not engage at intersections, traffic lights, or on city streets. Tesla's Robotaxi service (Texas) runs an unsupervised variant of the same FSD stack inside a geofence.
Which production EVs let me drive hands-free?
Five production systems permit hands-off operation today, all restricted to mapped divided highways: GM Super Cruise on ~750,000 miles of highway (Cadillac Lyriq / Vistiq / Escalade IQ; Chevy Silverado EV, Equinox EV, Blazer EV; GMC Sierra EV / Hummer EV), Ford BlueCruise on mapped 'Blue Zones' (Mustang Mach-E, F-150 Lightning, etc.), Nissan ProPILOT 2.0 on Nissan Ariya, Rivian Universal Hands-Free on the R1S / R1T (Gen 2), and Subaru Highway Hands-Free Assist on the 2026 Outback Touring. None of these handle intersections, traffic lights, stop signs, or city streets — they're hands-off lane-keepers on a defined road network. Tesla FSD operates on a far wider scope (including city streets and intersections) but is hands-on/eyes-on at all times because Tesla classifies it as supervised.
Can I buy a Level 3 (eyes-off) car in the US in 2026?
Not on a new car. Mercedes-Benz Drive Pilot was the only SAE Level 3 system certified for US public roads (California and Nevada, ≤40 mph on mapped highways), but Mercedes discontinued it for new models starting with the 2026 S-Class, replacing it with a hands-off/eyes-on Level 2++ system (MB Drive Assist Pro). Owners of 2023–2025 S-Class and EQS cars already equipped with Drive Pilot can still use it via subscription. BMW canceled its Germany-only Personal Pilot L3 program, and Honda Sensing Elite L3 remains Japan-only. So as of mid-2026 no new US vehicle is sold with an eyes-off L3 system.
What is the current status of Tesla's Robotaxi service?
As of August 2026, Tesla Robotaxi operates in 7 US metros: Austin, Dallas, Houston, Miami, Orlando, Tampa and San Francisco Bay Area. Rides are open and fully driverless across the Austin metro plus Miami, Orlando, and Tampa (Florida) — Tesla rolled out all three Florida cities driverless in July 2026 (Miami on July 3, then Orlando and Tampa on July 21), with Miami having been its first market outside Texas and California and its first day-one driverless launch — while Dallas and Houston are also open and the San Francisco Bay Area service still requires a safety driver because California regulates it as a limousine offering rather than a robotaxi. Rides are booked through the Tesla app, and the service runs an unsupervised variant of the same FSD stack behind Tesla's 1.48M active FSD subscriptions (Tesla's metric: includes upfront purchases, excludes free trials). Scale remains small: Texas filings authorize 42 vehicles, independent trackers put the active driverless fleet in the low tens, and Musk has tied aggressive scaling to the FSD v15 model rewrite expected late 2026 / early 2027 — though on the Q2 2026 earnings call Tesla said the robotaxi fleet is already running early v15 builds, with roughly 40% of 7 planned improvement tracks merged. Tesla's Q2 2026 disclosure also states it expanded the unsupervised operating area in Austin, and lists Phoenix, AZ and Las Vegas, NV as 'Preparations Underway' — Tesla-confirmed next markets rather than trade-press rumor, though no launch date was disclosed for either. Nevada has since acted: on July 27, 2026 the state's Transportation Authority granted Tesla a robotaxi permit but capped it at 10 vehicles on the Las Vegas Strip corridor, against the 5,000 Tesla requested. The purpose-built Cybercab is now in production at Giga Texas and began carrying employee riders on the Giga Texas campus in July 2026, but it is not yet in public Robotaxi service and Tesla has given no date for that.
What is the current status of Waymo?
As of August 2026, Waymo is the largest driverless robotaxi service in the US and the only one operating at real commercial scale: public, fully driverless rides in 11 US metros (Phoenix, San Francisco Bay Area, Los Angeles, Atlanta, Austin, Dallas, Houston, San Antonio, Orlando, Miami, Nashville), a fleet of roughly 3,500 vehicles, about 500,000 paid rides per week, and ~4M autonomous miles per week. Two things changed recently. Freeway rides, which Waymo paused in late May 2026, began returning on July 29, 2026 starting with Phoenix — surface-street service ran normally throughout the pause. And on July 8, 2026 Waymo announced four new US markets: Las Vegas, Denver, San Diego and Tampa. Those started employee-first (Las Vegas on July 8-9), with public service targeted by the end of 2026, so they are not yet counted among the live public cities above. London is also slated for later in 2026. Waymo has passed 20M+ lifetime paid trips and reaffirmed its goals of 20+ cities and 1M weekly trips.
What is the current status of Zoox?
As of August 2026, Zoox — Amazon's robotaxi arm — charges for rides on the Las Vegas Strip and runs a free service in San Francisco, using a purpose-built vehicle with no steering wheel or pedals. Because that vehicle has no manual driving controls, it needed an exemption from the Federal Motor Vehicle Safety Standards before it could charge, and on July 30, 2026 NHTSA granted one — the first commercial exemption of its kind for a purpose-built robotaxi, covering up to 2,500 vehicles annually for two years and waiving eight FMVSS. Paid service began in Las Vegas on August 10, 2026; the fare is a base charge plus distance and time, quoted in full before you confirm and honored even if the car takes a longer route, with destination fees at Harry Reid airport, Sphere and T-Mobile Arena. San Francisco is a different situation and is not next in line: anyone can download the app and join the waitlist there, but the rides are free because Zoox holds only CPUC Pilot permits, which prohibit charging fares, and it has said it has no plan to begin paid service in San Francisco. Austin is employee and friends-and-family access; Dallas and Phoenix are mapping with safety drivers and carry no passengers. Zoox reports nearly 2 million as of March 2026 (Zoox figure; no newer count published, and riders have since grown from ~350,000 to nearly 1 million, so the true figure is higher) and nearly 1 million riders across Las Vegas, San Francisco, Austin and Miami since the free service launched (Zoox figure, reported Aug 5, 2026); Las Vegas alone accounts for 500,000+ rides.
Tesla Robotaxi vs Waymo — how do they compare?
They're playing fundamentally different games. Waymo leads on driverless commercial scale today: 11 US metros, ~3,500 vehicles, ~500,000 paid public rides per week, and ~4M autonomous miles per week (freeway rides, paused since late May 2026, began returning on July 29, 2026 starting with Phoenix; surface streets ran normally throughout) — but it relies on HD-mapped cities and a lidar-heavy sensor stack, which is why expanding to a new city is a multi-quarter operation. Tesla's bet is different: a vision-only, generalized FSD stack that already runs (supervised) behind 1.48M active FSD subscriptions (Tesla's metric: includes upfront purchases, excludes free trials) and 10B+ cumulative miles, with the Robotaxi product live in 7 US metros (Austin, Dallas, Houston, Miami, Orlando, Tampa and San Francisco Bay Area) — unsupervised except the Bay Area, where California law still requires a safety driver. Tesla went driverless across Florida in July 2026 — Miami on July 3 (its first market outside Texas and California), then Orlando and Tampa on July 21 — each fully driverless from day one, but the active driverless fleet is small — independent counts through mid-2026 put it in the low tens, against 42 authorized in Texas state filings, and Musk has tied real scale to the FSD v15 model rewrite expected late 2026 / early 2027. Tesla asked Nevada for 5,000 robotaxis and got 10: the state's Transportation Authority approved its permit on July 27, 2026 but capped the fleet at ten vehicles on the Las Vegas Strip corridor, with no airport pickups. Today, Waymo has by far the larger commercial driverless service and fleet; Tesla has the larger consumer autonomy footprint and a faster city-add roadmap if the unsupervised stack proves out at fleet scale. Both can be true at once.
Are robotaxis safer than human drivers?
For exactly one operator there is now real independent evidence — and it does not transfer to the rest. The first independent per-mile crash benchmark of a driverless fleet was published by the Insurance Institute for Highway Safety (IIHS) on July 23, 2026. Measured per vehicle mile traveled against human drivers in the same cities and years, Waymo's rider-only vehicles were involved in 68% fewer police-reportable crashes overall, 85% fewer single-vehicle crashes, and 81% fewer injury crashes. The metro-level picture is uneven and worth stating plainly: 76% lower in Phoenix, 71% lower in Los Angeles, and 35% lower in San Francisco — but 4% HIGHER in Austin, where IIHS notes the sample was relatively small and draws no firm conclusion. Scope: roughly 50 million Waymo rider-only miles (2021-2024) compared against roughly 222 billion human-driven miles across those four metros. Method: built from NHTSA Standing General Order crash reports, with redundant and non-qualifying records culled and a police-reportability standard then applied — only about 22% of the 736 automation-engaged public-road crashes were judged likely police-reportable, and of Waymo's 89 police-reportable involvements 64 occurred in fully driverless operation. Three caveats are essential to reading this honestly: (1) Waymo is the only operator that publishes the vehicle-miles-traveled denominator, and it supplied that mileage voluntarily, so no other robotaxi operator can currently be benchmarked this way — this finding says nothing about Tesla Robotaxi, Zoox, or anyone else; (2) the Austin sample is small; and (3) IIHS itself cautions that the present data-collection system 'isn't good enough to allow continuous monitoring of a large-scale expansion.' IIHS is funded by the auto-insurance industry; there is no indication that Waymo funded the research. The honest summary: the best available independent research shows a large per-mile safety benefit for Waymo in three of the four metros studied, with Austin the exception, and that result cannot be extended to Tesla Robotaxi, Zoox, or any other operator, because none of them publish the per-mile denominator that would make the same comparison possible.
What happened to Cruise?
GM wound down Cruise robotaxi operations in December 2024, citing the cost and time required to scale a competitive robotaxi business. The decision is estimated to reduce GM spending by more than $1 billion per year. Cruise's technology and engineering talent were redirected into GM's personal-vehicle ADAS work, where Super Cruise continues to be developed.
Are there any Chinese robotaxis operating in the US?
No. Baidu Apollo Go (the global volume leader, which peaked above ~350,000 weekly driverless rides in March 2026), WeRide, Pony.ai, and AutoX all operate robotaxi services but none have a US public service. WeRide and Pony.ai have expanded internationally via Uber partnerships (Dubai, Abu Dhabi, Riyadh, Zagreb). AutoX took a California driverless testing permit back in 2020 but never launched a US service and no longer appears on the California DMV's current permit lists; WeRide's US presence is a California test permit only. The 100%+ US tariff on Chinese EVs effectively closes the door on direct vehicle imports.
Is the Uber robotaxi service real?
Uber doesn't operate its own robotaxis — instead it partners. In the US: Waymo is bookable through the Uber app in Austin and Atlanta. Uber's expanded deal with Lucid and Nuro now covers at least 35,000 Lucid vehicles (up from the 20,000 announced in July 2025), with Uber's total investment in Lucid reaching about $500M; Lucid builds the cars, Nuro supplies the Level 4 self-driving system, and first service is targeted for the San Francisco Bay Area later in 2026. Internationally: WeRide (Dubai/Abu Dhabi/Riyadh), Wayve (London, Tokyo with Nissan), and Pony.ai (Zagreb).
How much does a robotaxi ride cost?
Waymo prices comparably to UberX or slightly above in its service areas — typically a $2-5 premium per ride for the novelty + dedicated vehicle. Zoox began charging in Las Vegas on August 10, 2026, after NHTSA granted the exemption its steering-wheel-free vehicle needed to operate commercially (July 30, 2026, up to 2,500 vehicles annually for two years). The fare is a base charge plus distance and time, shown in full before you confirm, with destination fees at Harry Reid airport, Sphere and T-Mobile Arena — Zoox has not published a base fare, and says only that it aims to be competitive with ride-hail 'comfort' tiers. San Francisco stays free: Zoox holds CPUC Pilot permits there, which forbid charging for rides. Tesla Robotaxi in Austin has been priced as a flat $4.20 per ride at launch as a promotional rate, with regular pricing TBD. International operators charge in local currency at rates comparable to local ride-hail.
Can a robotaxi get a traffic ticket in California?
Starting July 1, 2026, yes. Under new California DMV rules (adopted April 29, 2026, implementing AB 1777), police can issue a 'notice of noncompliance' to the autonomous-vehicle company when a driverless car commits a moving violation — previously, with no human driver, officers could only write parking tickets. The AV operator is effectively treated as the driver and must report each notice to the DMV within 72 hours (24 hours for a serious incident like a crash). Repeat or serious violations can lead the DMV to restrict or suspend a company's permit.
What is SAE Level 4 vs Level 5?
Level 4 means no human driver required, but only within a defined operational design domain — e.g. specific cities, weather conditions, or speeds. Every robotaxi operating commercially today is L4. Level 5 means full automation everywhere, in any conditions — no production system exists at L5 in 2026 and no operator is publicly claiming a near-term L5 product.
Related on The Charge Port
Dig into the vehicles that ship these driver-assist systems, or jump to our other research tools.
Every operator status, city list, and capability claim on this page is reconciled against primary sources (operator press releases, SEC filings, state DMV/CPUC records, and dated trade reporting) on a monthly cycle. Inline footnotes link to the specific source for each claim. Where two sources conflict, we cite the most recent dated record from the operator or its regulator.
What we deliberately do NOT do: rely on speculative timelines from press tours or social-media announcements without a regulatory filing or live service to corroborate. Robotaxi rollouts have historically slipped by quarters or years; only operators with at least public-rideable service (or a filed permit) are listed as "Active" or "Limited" rather than "Preparing."
How we handle fleet numbers: operators rarely disclose them, so most published counts are third-party estimates —
and they routinely measure different things (total fleet vs. state-authorized vs. actively running vs. running unsupervised).
We state a precise figure only when it comes from a document (a regulator's permit, a state filing, an SEC filing or an operator's
own disclosure) or when two or more independent outlets agree on the same measured quantity. Where only one outlet has a number,
we keep the estimate broad or show the full spread rather than privilege a single count. This applies to every operator equally:
it is how we keep an outlet's editorial slant, in either direction, out of our numbers.
Last full verification: August 21, 2026. Refresh cycle: monthly.