Zoox Wins First US Approval for Steering-Wheel-Free Robotaxis

A century-old assumption finally bends

The US National Highway Traffic Safety Administration (NHTSA) has granted Amazon's autonomous driving subsidiary Zoox an exemption from parts of the Federal Motor Vehicle Safety Standards (FMVSS). The practical result: paid rides in a car with no steering wheel are now permitted in the United States for the first time.

Until now, nearly every robotaxi on American roads has been a retrofit. Waymo built on the Jaguar I-Pace, Cruise on the Chevrolet Bolt — production cars with the driver's seat, wheel, pedals and mirrors left intact, then loaded with sensors and compute. The reason is regulatory rather than technical. FMVSS presumes a human driver, and dozens of its clauses — pedal placement, instrument display, mirror field of view — simply do not parse without someone sitting up front looking forward.

Zoox threw that assumption out at the design stage. Its vehicle has facing bench seats, no driver's position, and a bidirectional layout with no meaningful front or rear. It was never a car that could pass the existing standards; it was a car that sat outside them entirely.

Why the 2,500-unit cap matters

Under the exemption, Zoox may operate up to 2,500 vehicles per year for the next two years. That ceiling tracks the volume limit US federal law allows for safety-standard exemptions. It is not a blank check — it is a compromise that lets regulators gather real-world data while capping downside exposure.

By fleet size, this is modest. Waymo alone already runs a comparable or larger fleet, and against the total US taxi and ride-hailing market the number is a rounding error. But the significance isn't measured in units. The precedent is the product. Any company now designing a purpose-built autonomous vehicle can start from "what conditions must we meet?" rather than "is this category of vehicle even legal here?" That shift in the opening question is worth more than a few thousand cars.

Reading the recall

Worth noting alongside the approval: Zoox recently recalled 105 vehicles and pushed a software update. The reported issue involved smoke detection software — a perception-stack limitation, not a mechanical fault.

The obvious objection writes itself: why approve a company that just recalled its fleet? A fair answer requires understanding how "recall" functions in this sector. These events are usually resolved over the air rather than in a service bay, and they are frequently self-identified and self-reported by the operator. Some regulators read that as evidence the oversight loop is working rather than failing. That said, it is not absolution. The real test of this decision will be how much incident, disengagement and intervention data NHTSA compels Zoox to publish once paying passengers are actually on board.

What Amazon is really buying

Amazon acquired Zoox in 2020, and it would be a mistake to read the company's interest as a straightforward bet on passenger transport. An autonomous driving stack is largely indifferent to whether the payload is a person or a parcel, and Amazon operates one of the world's largest logistics networks. Driver labor is a structural cost in urban last-mile delivery, and any technology that compresses it meets internal demand far larger than a taxi business would generate.

Seen that way, this approval is a move in the ride-hailing market and an early skirmish in logistics automation at the same time.

The view from Korea

South Korea runs its own framework — designated autonomous driving pilot zones under the country's autonomous vehicle act, with paid-ride demonstrations in Seoul districts such as Sangam and Gangnam. Most of those deployments, however, use modified conventional vehicles and typically require a safety operator on board.

The gap the US case exposes is less about technology than about regulatory tooling. NHTSA reached for a time-limited, volume-capped exemption to handle a vehicle that cannot satisfy rules written for a different era. Rewriting standards takes years; the exemption creates a channel for evidence to accumulate in the meantime. Korea will eventually face a purpose-built, wheel-free vehicle at its own door, and it is worth settling in advance how such a vehicle gets certified and how liability is apportioned when something goes wrong.

What could still go sideways

Optimism should be tempered. The robotaxi sector has already produced one spectacular failure in GM's shutdown of Cruise, and the industry learned that a single serious incident can flip public opinion and regulatory posture within weeks. The details that will matter next are unglamorous: the geofence Zoox actually operates in, the weather and speed envelopes it accepts, and whether NHTSA converts two years of exemption data into a genuine revision of the standards — or quietly lets the window close.