Wayve and Uber: London's first robotaxi runs without a map. It still runs with a driver.

On 3 September 2026, Wayve and Uber began carrying paying passengers in London. Riders opt in under Trip Preferences in the Uber app, and more than 140,000 Londoners had done so by launch day. If they are matched, a Ford Mustang Mach-E running the Wayve AI Driver collects them. It can go anywhere in London except the airports, at no extra cost, and the fare is shown upfront.

A trained, TfL-licensed private-hire driver sits behind the wheel “to oversee the trip”, in both companies' words. That sentence is the whole review in miniature. What launched in London is the most interesting driving software in Europe, carrying members of the public for the first time. What has not launched is a car that drives without a person.

Both of those are true, and the coverage has mostly reported one or the other. This review tries to hold both.

We have already reviewed this car

The Ford Mustang Mach-E is on this site as a consumer car. Its selling point is BlueCruise, which lets the driver take their hands off the wheel — but only on 130,000 miles of pre-mapped divided highway that Ford calls Blue Zones. Off the map, the hands go back on.

Wayve's Mach-E inverts that exactly. It needs no map: the software is meant to drive any street it has never seen. But it cannot let the human go. The consumer car trades a map for a free pair of hands; the robotaxi has done without the map and kept the human. Same body, same battery, different software — and the difference between them is the most important open question in the industry.

What “mapless” means, and why it matters to this site

Most driverless systems, Waymo's included, lean on a high-definition map of every road they operate on: lane geometry, kerbs, signals, surveyed to centimetres in advance. The car compares what it sees with what it expects. That works, but every new city has to be surveyed first, and every change to the road leaves the map out of date.

Wayve calls its approach AV2.0: an AI Driver that “learns from experience, much like a human, allowing it to scale and adapt to complex environments without relying on traditional HD maps or hand-coded rules.” One neural network, trained on driving data, is asked to generalise. The company says it tested the system in more than 500 cities across Europe, North America and Japan in a single year, “zero-shot” — without training specifically for each one.

This matters here because both of this site's in-depth pieces ended in the same place. The export gap tested every explanation for why Chinese cars lose their navigation assist abroad, and mapping was one of only two that survived. The motorway paradox found that Waymo's freeway recall this summer came down to roadworks: a road that changed overnight and a map that did not know. If Wayve's claim holds, it removes the barrier both pieces identified. That makes it the most consequential claim anyone in autonomy is making.

London does not test that claim yet. A supervised launch shows the software can carry the public safely while a professional watches, which is real progress. It cannot show whether mapless driving is safe enough to remove the professional — and that is the only version of the claim that changes anything.

The sensor rack, head-on, above the windscreen of an otherwise standard Mustang Mach-E. Wayve's launch material describes the fit only as “surround sensors”, so our comparison tool lists the hardware as undisclosed. Source: Wayve

What the law says it is

Legally, what launched on 3 September is private hire. Tech Times reports that Transport for London licensed the cars as private-hire vehicles, not as automated vehicles, under a supervised-pilot framework the government accelerated in May 2026.

Taking the driver out is a different process under the Automated Vehicles Act 2024. According to the same reporting, which cites analysis by the law firm Jones Day, an operator needs three things in sequence: the vehicle authorised as self-driving by the Vehicle Certification Agency, a Vehicle Special Order, and an automated passenger services permit. As of launch day, Tech Times reports, no operator had completed even the first step, and Jones Day's assessment puts driverless approval in 2027 at the earliest. Wayve's chief executive, Alex Kendall, has said the company is pushing as fast as it can, and has not given a date.

Two definitions of “first”

London now has two claims to its first robotaxi, and they mean different things.

Wayve was first to carry the public — with a driver aboard. Waymo has been testing about 100 Jaguar I-Pace vehicles across a 100-square-mile area of London since April 2026, also with safety operators, and has said it intends to launch rider-only, with nobody in the front. On current evidence, Waymo is likely to be first to remove the driver, and Wayve was first to take a fare.

They are also running different experiments. Waymo is bringing a mapped system with a decade of driverless operation in American cities to a new, harder city. Wayve is bringing an unmapped system to its home city with no driverless record anywhere. If both reach driver-out, London becomes the first place where the two philosophies can be compared street for street.

The business underneath

The robotaxi is Wayve's shop window, not its whole business. In February 2026 it raised $1.5 billion at an $8.6 billion valuation from a list that reads like a car industry directory: Microsoft, Nvidia, Uber, Mercedes-Benz, Nissan and Stellantis, alongside SoftBank. Its stated market is “every vehicle”, and its route there is licensing the AI Driver to carmakers for their own assistance systems. Wayve says Mercedes-Benz has since signed a production agreement for select future models, and Nissan has agreed a robotaxi pilot in Tokyo with Wayve and Uber.

That makes London a demonstration to two audiences at once: riders, and the carmakers deciding whether to put Wayve's software in millions of private cars. Tech Times reports that $300 million of Uber's commitment is tied to the London deployment. Neither company has published those terms.

What we cannot tell you

• How often the driver takes over. Wayve has published no disengagement data for London. Until it does, “supervised” could mean anything from an idle passenger to a busy co-driver.
• The fleet size. The companies say “a small number”. The figure of 15 is reported, not announced.
• A driverless date. None has been given, and the legal route has three steps that no operator had begun at launch.
• A safety record. There is no public record of Wayve operating without a human anywhere, so there is nothing yet to compare with Waymo's.