Taking a ride in London's first autonomous taxi

“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.”

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.

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.