The U.S. government has given Zoox, a self-driving car company owned by Amazon, permission to start charging passengers for rides in its robotaxi. The vehicle has no steering wheel, no pedals, and no traditional controls a human driver would use. The approval was published in the Federal Register on July 30, 2026. TechCrunch

Zoox CEO Aicha Evans said this is the first time the government has granted this kind of commercial permission for a purpose-built robotaxi. The permission limits Zoox to 2,500 vehicles per year for two years. It covers several safety rules that were written with human-driven cars in mind, including ones about windshield defrosting and braking systems. The government also said Zoox can test its automatic emergency braking at the vehicle's top speed rather than at a fixed 90 mph standard. Federal Register

A Zoox spokesperson told TechCrunch the company will start charging for rides first in Las Vegas, with other cities to follow. California is not yet on the list: Zoox still needs separate permits from the state Public Utilities Commission and the Department of Motor Vehicles before it can charge passengers there. TechCrunch

The road to this approval began in August 2025, when the government allowed Zoox to operate on public roads and give free rides but not charge for them. That earlier approval was described as the first for U.S.-made self-driving vehicles. NHTSA It came with some trouble: Zoox agreed to recall 270 driverless vehicles after an incident involving an unoccupied robotaxi. Reuters

The paperwork leading to this week's approval went through several stages. The government published a notice that Zoox had applied for the exemption on September 25, 2025. Federal Register A follow-up notice on March 11, 2026 showed the review was ongoing. Federal Register The government asked for public comment on the application in March 2026. Reuters

The limit of 2,500 vehicles per year for two years is not unique to Zoox. It comes from an existing law that lets manufacturers deploy a small number of vehicles that do not meet standard safety rules, mainly to test new technology. NHTSA referenced this same framework as far back as a March 2022 notice about safety standards for self-driving vehicles. Federal Register

On the same day as the Zoox announcement, the government made two broader moves. NHTSA updated its exemption process to make it easier for automakers to temporarily sell a limited number of non-standard vehicles for testing. NHTSA Administrator Jonathan Morrison said the agency supports the safe development of self-driving vehicles and is taking a balanced approach. NHTSA Separately, NHTSA announced a partnership with SAE Industry Technologies Consortia to fund a three-year, $5 million effort to gather data and speed up the creation of safety standards for self-driving vehicles. TechCrunch NHTSA is also reviewing a similar application from a Los Angeles startup called Robomar, which suggests more companies are lining up for the same kind of permission. TechCrunch

NHTSA said Zoox will face an oversight structure that can adapt as its technology improves. TechCrunch The agency did not provide specifics about enforcement or how often Zoox must report, though the earlier free-ride approval already required Zoox to cooperate with a recall investigation.

The broader context here is that these safety standards were written decades ago for cars that humans drive. A rule about windshield defrosting, for example, assumes a human driver needs a clear view through the front windshield. Zoox's vehicle is designed to drive in either direction and has no traditional front, so that rule does not really apply. The braking standard had a similar issue: the government recognized that testing a self-driving vehicle's braking at 90 mph makes little sense if the car cannot go that fast.

What stands out is that NHTSA appears to be building a general system rather than handling each company one at a time. The Zoox approval settles one petition, but the updated process and the new consortium suggest the agency wants a framework that works for all self-driving vehicles. The Robomar application waiting in the queue will be an early test of whether that broader approach works.

In this author's view, the 2,500-vehicle limit and the two-year window suggest the government wants to see how these vehicles perform in real commercial use before writing permanent rules. That is a cautious approach, and arguably the right one given the safety concerns at stake. The talk of adaptable oversight, while vague, at least acknowledges that rules written for human-driven cars cannot stay frozen while the technology underneath them keeps changing. The open question is whether the government can write new standards quickly enough to keep up with a service that, starting in Las Vegas, is now live and charging passengers.

For the self-driving car industry, the Zoox approval shows that the federal pathway works, at least for a well-funded company willing to go through a multi-year approval process. Whether that pathway can handle more than one company operating under a vehicle cap is the question the government's new consortium is meant to answer.