Check state DMV, public utility and transportation records, plus the operator's own map, to test a robotaxi launch claim before you believe it.
How do you verify a robotaxi deployment announcement?
Treat the announcement as a claim, then look for three independent records: a state authorization, a defined service area and a booking path that works. A press release can describe a launch before any passenger can hail a ride. Regulators publish what a company is allowed to do and where.
NHTSA says states permit a limited number of self-driving vehicles for testing, research and pilot programs, and that no fully automated vehicle is for sale to consumers (NHTSA). A robotaxi is a permitted service, not a product you buy. The rider side is explained in how robotaxis work.
Testing, deployment and paid service are different permissions
California separates them. The DMV authorizes three categories: testing with a safety driver, driverless testing, and deployment. Each category has approved locations, operating conditions and speed limits (DMV).
A vehicle approval is not a taxi license. Paid passenger service needs a second permission from the California Public Utilities Commission. The CPUC runs pilot programs, where monetary compensation may not be charged, and deployment programs, where fares are allowed (CPUC).
| Permission | What it allows | Where to verify |
|---|---|---|
| DMV testing with a driver | Testing on any public road in California with a safety driver | DMV permit holder list |
| DMV driverless testing | Testing without a driver in listed places and conditions | DMV driverless table |
| DMV deployment | Deploying vehicles in listed places and conditions | DMV deployment table |
| CPUC driverless deployment | Paid passenger service without a safety driver | CPUC permit table and ODD maps |
| TxDMV authorization | Commercial automated vehicle operation in Texas | TxDMV report and status lookup |
Check the state vehicle regulator first
Start with the DMV’s permit page. As of October 23, 2026, it listed testing permits with a driver for 26 companies. It listed driverless testing for six: AutoMo, Nuro, Tensor, Waymo, WeRide and Zoox.
Each driverless row gives approved cities, hours and an operational design domain. The deployment table is shorter. As of October 22, 2026, it listed Mercedes-Benz Research and Development North America, Nuro and Waymo.
Mercedes’ deployment row covers freeways in several regions at speeds up to 40 mph. That is a vehicle approval, not evidence of a taxi stand on your street. Read the columns, not just the company name.
A testing row can put a logo in a release without passenger service. A deployment row can cover a freeway feature rather than a hailed ride. The permit type is the fact that matters.
Check the passenger-service regulator next
The CPUC’s permit table lists a permit type for each carrier and, for driverless deployment holders, an operational design domain statement and maps.
On November 8, 2026, the table showed one driverless deployment permit, held by Waymo. Driverless pilot permits were listed for Waymo, WeRide and Zoox, and drivered pilot permits for Nuro, Tensor, Waymo and Zoox.
Pilot means no fare. That single word separates a demonstration from a commercial service. The CPUC also bars passenger service at airports without the airport authority’s authorization, and posts those notices separately.
Use a second state to test the claim
California is not the only model. Texas created a required authorization for commercial automated vehicle operation in Senate Bill 2807 in 2025. The requirement became enforceable on May 28, 2026 (TxDMV).
TxDMV publishes an authorization report. The version dated October 29, 2026, listed 17 companies and 3,164 automated vehicles. Waymo accounted for 1,259 and Tesla Robotaxi for 1,080.
Those are authorized vehicles, not passenger trips. The report has no city map or service-hour data, so it cannot prove a pickup at a given address.
What crash data can and cannot prove
NHTSA’s Standing General Order requires named operators to report certain crashes. For an automated driving system, a report is due when the system was in use within 30 seconds of a crash involving a fatality, a vulnerable road user, an air bag deployment, a tow-away or a hospital trip.
The third amended order took effect on June 16, 2025. The most severe reports are due within five days, and less severe ones monthly. NHTSA publishes the data as CSV files.
The agency warns that the data is not adjusted for fleet size or miles traveled. Operators do not submit exposure figures. A higher count can reflect better telemetry, not worse safety.
Finish with the company’s own service map
The operator’s app is the last check, because it shows what you can book. Enter a real pickup and drop-off pair inside the advertised city. A city name on a press release can mean a waitlist, a testing permit or a single neighborhood.
Separate an “up next” list from live service, and check the hours and the quoted fare before you confirm. If the app will not offer a trip, the service is not available to you yet, whatever the announcement says.
The Automotive News and Industry hub carries more launch coverage. For a parallel checklist on technical claims, see evaluating charging-speed records.
Frequently asked questions
Does a state testing permit mean I can hail a robotaxi?
No. A testing permit allows a company to operate under listed conditions, often without passengers or fares. Paid passenger service needs a separate authorization in states such as California.
Is a driverless deployment permit enough to charge fares in California?
No. The DMV deployment permit covers the vehicle and its operating conditions. The CPUC permit covers passenger service and fares. A company needs both.
Do all states publish a robotaxi permit list?
No. Requirements differ. California publishes detailed permit tables. Texas began requiring commercial authorization in 2026 and publishes an authorization report. Check the state regulator that covers that city.
Can NHTSA crash reports tell me whether a robotaxi service is safe?
Not directly. The reports cover specific crash types and are not adjusted for miles, fleet size or operating area. They can confirm that vehicles were operating, but not rank safety between companies.
