A ride in UK's first autonomous taxi reveals the future of driverless cars

Close

News

5 mins read

17 November 2025

The Wayve Ford Mustang Mach-E has been cruising successfully hands-free through north London traffic for about 40 minutes and now faces a tough test: an unprotected right turn into a busy street. A parked black cab spoils the view on the left while the usual broad spectrum of London vehicular traffic dashes on by, not giving an inch. But the Ford behaves remarkably humanly, creeping out until it spots a gap and then nips into the flow without drama.

The demonstration is a powerful case-maker for what UK firm Wayve calls AV 2.0, where artificial intelligence (AI) makes decisions rather than the car following a pre-programmed set of rules.

Robotaxi 1.0 in this definition is Google’s Waymo, the pioneering autonomous service successfully operating in five US cities with a claimed 10 million rides under its belt.

Waymo's 2000-strong Jaguar I-Pace fleet is instantly recognisable and has done much to normalise being ferried around without a driver, with Waymo actually overtaking Uber rival Lyft in terms of share of the taxi market in San Francisco.

But the huge bank of sensors that makes those Jaguars so recognisable is also the hurdle to wider acceptance – they cost a lot of money. Waymo hasn’t given any indication of what the bill really is to convert an I-Pace into a robot, but estimates start around $30,000, due to multiple cameras and radar and lidar units.

Wayve still has a sensor rig on its test Fords but claims it has drastically reduced that hardware bill to between $1000 and $2000.

“We’re seeing a paradigm shift here from AV 1.0 to AV 2.0,” Alex Kendall, co-founder and CEO of Wayve, told Autocar.

However, Wayve’s next-level approach to hands-free driving is far more than about stripping cost from sensor set.

“A lot of the systems driving in Shanghai or San Francisco do so with a very complex set of rules that takes an army of engineers to hand-code and typically doesn't deal with the complexity of an environment,” Kendall said.

Wayve guides cars using ‘end-to-end’ AI, which makes decisions on the fly based on learned intelligence only. This ‘generalisation’ is better at dealing with the unfamiliar, so Wayve claims.

“You're never going to see everything in your training data,” Kendall said, “so it's all about being able to learn general concepts and be able to apply those concepts to new things.”

Back to top

Whereas Waymo cars are carefully geofenced to operate in a specific area mapped in high definition, Wayve claims to able to operate from the get-go in an unfamiliar city.

“We can go to any traffic light in the UK, Europe, Asia or North America and the car can drive through it,” Kendall said.

Wayve is siding with Tesla and its Robotaxi programme, which also uses end-to-end AI, albeit with more reliance on sensors than Tesla. The problem, however, will come with regulators.

“Regulators in Europe will not accept end-to-end AI,” Christian Senger, CEO of Volkswagen Autonomous Mobility, told Autocar in September. “You can't launch a fully end-to-end AI service where you can't prove how this system really works.”

Volkswagen is accelerating its own robotaxi programme using Moia-branded ID Buzz vans loaded with Mobileye’s self-driving system, which operates both AI-led and rules-based ‘deterministic’ autonomous systems.

The Moia programme will move to the US in 2026 using 10,000 ID Buzz vans operated by Uber. Currently the manufacturer operates a trial scheme in Hamburg, Germany. Next year it hopes to reach the point when it can tell regulators "we’re better than a human", according to Senger.

Back to top

Moia is banking on winning around regulators not just with a more transparent decision-making system but a complete service.

“Everybody has a fragment of a product,” Senger said of the likes of Wayve, Rimac-backed Verne and other autonomous driving companies. “Imagine you were in charge of a city’s mobility administration. How to start [licencing robotaxi companies]? Who is liable? What kind of education would my staff need? What kind of facilities do I need? Moia would offer a turn-key solution,” Senger said.

He cites examples such as exiting a parking garage when there’s no human hand to take a ticket or knowing where to legally drop off a customer.

Wayve isn't concerned with operating robotaxis; it just wants to sell its AI Driver software to anyone willing to integrate it into a vehicle. So far Nissan is the first to announced it will incorporate it into a Japanese-built production car starting 2026, but Kendall said more manufacturers are poised to sign up.

Nissan’s integration won’t be autonomous but instead what’s described as level two plus, meaning hands off but eyes on. But Nissan will build in a lidar unit to give it opportunity to upgrade to level three, meaning eyes-off, hands-off motorway driving, at a later date.

For Wayve, the leap from level two plus to level four isn't so big. My London demo was level two plus, given the presence of the safety driver at the wheel, but given that he didn’t touch the steering wheel or pedals once in the near hour-long trip, a robotaxi application can easily be imagined.

Other than Volkswagen and Tesla, car makers have been wary of pouring money into a technology that so far has proven expensive, dependent on regulators and not scaleable. Both Ford and General Motors in the US have backed away from their AV programmes, despite the more lax regulatory environment across the Atlantic.

There are signs, however, that the ‘trough’ period after the initial hype is returning to cautious optimism. All eyes are on Tesla and whether CEO Elon Musk’s promise a 1500-strong driverless robotaxi fleet materialises by the end of the year.

Back to top

“Their success would be extremely valuable for us to monitor, because that will tell us what we can do and what we should not do either,” Simone Fabris, Wayve’s head of product and delivery, told Autocar.

If Wayve could sell car makers a system with robotaxi capabilities, the rush could start again.

“Those OEMs that do not pursue the ADAS robotaxi experience will ultimately run the severe risk of running behind in this roborace,” José Asemundi, head of global automotive research at the bank JP Morgan, wrote in an July report.

Chip making giant Nvidia is helping the robotaxi push with its so-called AGX Hyperion 10 chip and sensor architecture, which it claims “makes any vehicle level four-ready”. It has signed Stellantis, Lucid and Mercedes-Benz, along with autonomous development companies such as Wayve and Chinese rivals including Pony.ai and Momenta.

Others car makers are pushing robotaxis too, including Xpeng and Lotus.

The promise of robotaxis has always been tantalisingly out of reach, with public failures threatening to halt operations at any paint, as happened with GM’s Cruise. Waymo is currently dealing with public anger at the killing of a favourite neighbourhood cat in San Francisco.

However, if regulators can be persuaded that AI-based systems offer a leap forward in safety, the jump from techno dream to reality in Europe could be close at hand.