Uber and Wayve start London robotaxi rides with safety drivers still aboard

Uber and Wayve began carrying passengers in London on 3 September in what the companies describe as the city's first robotaxi service. Fifteen autonomous vehicles are licensed to operate, riders who request UberX, Uber Electric or Uber Comfort may be matched to one at no extra cost, and the service covers the city apart from the airports. A human supervisor sits behind the wheel of each vehicle and can take the controls. That last detail is the one to hold on to. A service with a person in the driver's seat who can grab the wheel or hit the brakes is a supervised trial operating commercially, not a driverless taxi. Under the UK government's pilot framework, operators including Uber and Wayve still need approval to run without a human safety driver, and neither company has it. The technology may or may not be ready. The permission is what is missing. The vehicles are Ford Mustang Mach-E electric cars fitted with Wayve's driving system, which reads the road in real time using roof-mounted sensors. Wayve's technical position has always been distinct from most of the field. Rather than relying on high-definition prior maps of every street, its approach is built around learning to drive from data in a way intended to generalise to roads the system has not seen. If that holds up, the cost of entering a new city falls sharply, because the expensive part of conventional autonomy is the mapping and validation work that has to be redone everywhere. London is a demanding place to test that claim. The road network is irregular, lane discipline is loose, cyclists and buses share space aggressively, and the weather is exactly the kind that camera-led systems find hardest. Choosing it as a launch city is a statement about confidence in generalisation, and the presence of a safety driver is the acknowledgement that the statement is not yet proven. Fifteen vehicles is a small number, and deliberately so. The commercial significance is not the fleet, it is the integration. The vehicles sit inside the ordinary Uber app, matched to standard product tiers, at standard fares. Riders are not opting into an experiment through a separate service, they are being matched to one. That is the model Uber has been building toward with every autonomy partner: stay the demand layer, let others own the driving stack, and add capacity to the network without owning it. Wayve has said it plans to expand next to Tokyo, where it is working with Uber and Nissan, and then to a further ten cities with Uber. Announced expansion plans in autonomy have a poor record of matching their timetables, and the London deployment does not yet demonstrate the thing that would make those plans credible, which is a system running without a supervisor. The competitive context in London is about to get more interesting. Waymo has said it plans to launch in the city, which would put the most operationally experienced driverless operator in the world into the same streets as the most prominent proponent of map-light, learned driving. If both run there, London becomes the first major market where the two approaches can be compared on the same roads under the same regulator. What has not been published is the operational detail that would let anyone judge progress. Neither company has released intervention or disengagement rates for the London vehicles, the criteria and timetable for removing the safety driver, the specific streets or manoeuvres excluded from the operating domain, or how many vehicles they intend to have running by any given date. Nor is it clear what regulatory milestone would trigger a move to unsupervised operation. The next real milestone is therefore not a bigger fleet. It is the first London vehicle carrying a passenger with the driver's seat empty, and the approval that allows it.