What TfL actually licensed

Transport for London issued Private Hire Vehicle licences covering up to 15 modified Ford Mustang Mach-E vehicles, fitted with Wayve's AI Driver software, surround-view cameras and radar, to carry fare-paying passengers booked through Uber. Wayve and TfL describe the approval as resting on a triple lock: the operator (Uber), the human driver, and the vehicle itself must each independently hold a TfL licence before any autonomous trip can run. That is a materially different bar from a software type-approval or a limited testing permit, because it licenses the commercial service end to end, not just the technology.

The trial operates under the UK government's AV Trialling Code of Practice, the framework London has used to bring supervised autonomous vehicles onto public roads while full driverless legislation, the Automated Vehicles Act, is still being implemented in stages. Each Private Hire Vehicle licence TfL grants under this scheme runs for one year, and TfL has stated it will keep monitoring the vehicles and can act if passenger safety is compromised, which keeps this a conditional, renewable approval rather than an open-ended one.

Supervised is doing a lot of work in that sentence

A trained, TfL-licensed private hire driver must remain in every vehicle, ready to take over if needed. That single requirement is the whole regulatory and commercial story here: it is why Uber can book the cars today under its existing private hire operator licence, and it is why the insurance question is simpler than it will be once the driver is removed. With a licensed human driver present and legally responsible for the trip, standard private hire liability and insurance frameworks still largely apply, even though the vehicle is doing the driving most of the time.

That will not hold once TfL or a future operator applies to run these vehicles without anyone in the driver's seat. At that point liability shifts toward the vehicle operator and the software maker in a way UK insurers have not yet priced at scale, and the Automated Vehicles Act's liability provisions, which shift responsibility for a self-driving vehicle's actions away from the human occupant, only take effect for genuinely driverless operation. Any owner of a mobility, delivery or logistics fleet watching this space should track that specific transition, supervised to unsupervised, rather than the robotaxi headline itself, because that is the point where today's insurance model stops applying.

A licensing model built differently from how Waymo got there

In the United States, autonomous vehicle operators have had to win approval city by city and state by state, through bodies such as California's Department of Motor Vehicles and Public Utilities Commission, with no single national licence covering the whole country. That patchwork is also where the model's biggest failure happened: California suspended Cruise's driverless permit in 2023 after one of its robotaxis dragged a pedestrian, a decision made entirely at state level with no federal body involved.

London's approach concentrates the same three checks, operator, driver, vehicle, inside one city regulator, TfL, operating under a single national Code of Practice rather than a state-by-state patchwork. It does not remove the risk of a serious incident forcing a suspension, TfL has said explicitly it can act if safety is compromised, but it does mean any UK or EU city weighing the same question is looking at one coherent national framework to plug into, not fifty different state regimes to negotiate separately.

What other EU and UK cities should watch next

London is the first UK city to reach this stage, and its result becomes the reference case every other European transport authority now has to react to. A city government deciding whether to let an autonomous ride-hailing pilot onto its streets can point to a concrete precedent: a triple-lock PHV licence, a mandatory safety driver, a one-year renewable term, and an explicit regulator override if something goes wrong, rather than a hypothetical framework.

The honest reading is that London has not solved the driverless question, it has built the supervised on-ramp to it, and the next real test is not this summer's limited rides but whether TfL is willing to license the same fifteen cars without anyone in the driver's seat once safety data from the supervised phase accumulates. That decision, not this one, is the one worth watching.