London’s first self-driving taxi service launched on 3 September, with rides in Wayve’s AI-powered vehicles on the Uber platform beating Waymo and Baidu to market in the city. The fleet is small: only 15 vehicles are currently licensed, with each still carrying a trained safety driver behind the wheel, against more than 100,000 traditional human-driven taxis already operating across the capital.

As with other Uber robotaxi deployments, autonomous lifts are hailed in the same way as any other Uber ride, albeit specifically through UberX, Comfort or Electric. They will be priced at standard fares with no surcharge and can travel anywhere in London except to the airports. Customers will essentially be matched at random with the vehicles, although the odds of such a pairing are very low given the current fleet size.

Safety appears to be at least on par with Tesla’s Austin robotaxi service. During a Bloomberg test drive, the safety driver took over twice in roughly 40 minutes: once to avoid blocking oncoming traffic while merging into gridlock and once to change lanes around stationary vehicles, with hands hovering just below the wheel the rest of the time.

Wayve’s underlying approach differs deliberately from most rivals. Its vehicles use cameras and radar rather than LiDAR, hewing more closely to Tesla’s camera vision-only approach than players like Waymo and Baidu. However, the technology also relies on a single AI model trained to generalise across unfamiliar roads rather than pre-built HD maps or geofenced zones. This is a choice the company says lets it handle London’s irregular street layout without the operational limits that constrain competitors elsewhere.

However, the gap between the 3 September launch and genuine driverless operation remains quite wide. Transport for London (TfL) confirmed that no operator, Wayve included, has even completed the first formal step toward unsupervised service, registering vehicles with the Vehicle Certification Agency, and a Driver and Vehicle Standards Agency (DVSA) permit for fully autonomous operation appears quite unlikely to arrive in 2026. Wayve co-founder Alex Kendall declined to give a timeline for removing the safety driver to The Guardian, saying only “we’re pushing as fast as we can”.

Uber’s approach to London reflects a far broader, deliberately non-exclusive strategy rather than a single bet on Wayve. The company is trying to catch as many major self-driving players in its net as possible and has committed more than US$10bn across partnerships with Wayve, Waymo, Zoox, Baidu, Nuro and several others. 

However, Uber’s relationship with Waymo has grown visibly strained in recent months, occasionally in public ways: Uber’s own chief technology officer publicly called a viral Waymo incident “scary” earlier in 2026. In the background, the two are in the process of winding down their deployment partnership for Phoenix and Atlanta in the US as Waymo looks to venture out on its own through its exclusive app. The latter is expected to launch its own London service by year-end but has said it will not do so with a safety driver present, putting the two companies on directly competing terms in the same city with no shared commercial arrangement here.

That expansive autonomous driving strategy is being tempered in places: Uber has allied with driver unions to slow robotaxi deployment in states including New Jersey and Washington DC. It is pushing for rules that would require human drivers to provide a minimum share of rides on any platform offering autonomous service. Uber President Andrew Macdonald has acknowledged the position “might feel ironic” given the company’s history, and it arrives as Uber cuts roughly 3,300 jobs and faces investor pressure to clarify its self-driving strategy after selling its in-house autonomous unit in 2020.

Beating Waymo to London matters more for perception than substance, since both companies still require a human behind the wheel. The launch’s real test will not be determined until TfL, the Vehicle Certification Agency, and the DVSA move past the registration stage neither operator has yet reached.