A Staged Debut, Not a Shortcut
On August 5, 2026, Transport for London granted Uber and Wayve private hire vehicle licences covering up to 15 autonomous Ford Mustang Mach-E vehicles fitted with Wayve's AI Driver software, cameras, and radar (Zag Daily). The grant completes London's "triple-lock" system, under which the operator, the driver, and the vehicle must all hold licences from the same authority (Wayve). Crucially, a trained, TfL-licensed private hire driver still sits in every car, ready to take the wheel — this is supervised autonomy, not driverless dispatch. More than 100,000 Londoners registered interest in riding one over the preceding eight weeks (Tech Digest).
What the licence does not do is authorise fully driverless trips. For that, Uber and Wayve need a separate Automated Passenger Service (APS) permit from the Driver and Vehicle Standards Agency, created under Part 5 of the Automated Vehicles Act 2024 — and neither company has one yet (Zag Daily).
Two Permits, Two Risk Profiles
The APS regime is a genuinely separate regulatory track from ordinary taxi licensing, not a rubber stamp. Permits are issued nationally by the Secretary of State for Transport via the DVSA, but a local licensing authority such as TfL must first give consent — deemed granted if it doesn't respond within six weeks (Burges Salmon). Once granted, a permit disapplies the ordinary taxi and PHV rules for that operator, replacing them with a bespoke safety regime: a full Safety Management System covering software updates and cybersecurity, immediate notification of safety-related incidents, short-term written reports within 30 days, and ongoing performance reporting to the regulator (gov.uk; Burges Salmon). Following its April 2026 consultation response, the government also widened the scheme so non-safety incident data can be shared directly with emergency responders, so paramedics and fire crews know how to approach a vehicle with no human driver (gov.uk).
Steelmanning the Extra Layer
There's a real case for treating the two permits as distinct. A supervised vehicle has a licensed, insured, accountable human in the seat who can intervene if the software misjudges a cyclist or a sudden road closure — the existing PHV framework, built around driver conduct and vehicle roadworthiness, covers that risk adequately. Remove the driver, and the entire liability and safety chain shifts onto the software stack and the operator's engineering practices, which conventional taxi law was never built to audit. Requiring a dedicated safety case, incident-reporting cadence, and data-sharing infrastructure before that handover is not bureaucratic caution for its own sake; it's the minimum evidentiary base a regulator needs to justify removing the human backstop at all. GMB's regional organiser Matthew Wright captured the constituency worried about a rushed transition, warning TfL that "driverless cars cannot replace the knowledge, experience, and human connection that a skilled driver offers" and urging "extreme caution" (Tech Digest). Whatever one thinks of the labour argument, the underlying safety instinct — don't let go of the wheel until you can show your homework — is sound.
Why the Split Is Good Design, Not Drag
Where this framework earns credit is in sequencing. Rather than force companies to choose between a blanket ban and an all-at-once driverless launch, the UK lets Uber and Wayve commercialise supervised rides immediately, generating real operating data, while the DVSA permit — with its heavier evidentiary bar — develops in parallel. That's a more coherent path than an outright moratorium, and it avoids the trap of writing driverless rules in the abstract, disconnected from how the vehicles actually perform on London's roads. The risk to watch is not that the APS permit exists, but that it becomes a bottleneck: if DVSA processing drags well into 2027 while the supervised fleet quietly proves itself safe, the second permit stops functioning as evidence-gathering and starts functioning as pure delay. Regulators should treat the accumulating supervised-ride data as an input to a faster APS decision, not as a reason to keep drivers in seats indefinitely once the safety case is made.
What Comes Next
The real test is whether DVSA can turn safety-case review into a predictable, time-bound process once Uber and Wayve apply for an APS permit — and whether TfL exercises its six-week consent window as a genuine check rather than a veto. A regime that can say yes to supervised rides in August and no to driverless ones in the same breath, for defensible and separable reasons, is the kind of proportionate regulation this publication generally wants to see more of. The next marker to watch is the date of the first APS permit ever granted under the 2026 regulations — that will show whether the two-tier system accelerates deployment or just adds a second queue.