The three listed problems all resolve to "Ignore" or reference text not present in the article body. None of the named quotes appear as claims in the article as written by the author (they exist only in the evidence pack). Therefore no sentences in the article require editing.
"For the vehicles to be used without a human driver, a separate approval would be needed from the Driver and Vehicle Standards Agency."
That single sentence, buried near the bottom of the BBC's write-up, is the whole story. The headline above it reads "Robotaxis granted licence to operate in London." The fine print says a qualified human must sit behind the wheel and remain legally responsible for the car.
The headline says robotaxi. The fine print says a qualified human must sit behind the wheel and remain legally responsible for the car.
Those two statements describe different things. One is a robotaxi launch. The other is what actually happened on 5 August 2026.
What Transport for London granted Wayve is a Private Hire Vehicle licence — the same class of paperwork a minicab needs. It completes what Uber calls the "triple-lock" requirement: operator, driver, and vehicle must all hold licences from the same authority. Wayve's all-electric Ford Mustang Mach-Es passed inspection, so the third lock clicked shut.
Read the triple-lock closely and notice what it contains. A driver lock. There is a licensed human in the definition of the thing being celebrated.
The trips will run under the government's AV Trialling Code of Practice, using Uber's existing TfL Private Hire Operator licence. Wayve's own language is careful: the vehicles "will do the driving, with a trained and TfL licensed private hire driver onboard to oversee the trip and provide support or take over driving if needed."
That is a supervised trial dressed in the vocabulary of arrival. The Verge said it plainly — "a supervisor still needs to be aboard each vehicle ready to intervene if needed." TfL was blunter still: the drivers "will still be in charge and have to take full responsibility."
None of this is a knock on the engineering. It is a knock on the scoreboard everyone is using to keep score.
Because here is the tension worth sitting with. Wayve's entire pitch — AV2.0, the "AI-first" approach — is that it needs no HD maps, no hand-coded rules, no geofenced domains. It learns to drive from experience, the company says, adapting to new roads and cities "with unprecedented speed and efficiency."
That is a claim about generalization. A claim that the system does not need to be spoon-fed a bounded operating domain to behave safely.
And yet the regulatory bar it just cleared is the one that keeps a human ready to grab the wheel. The most map-free, geofence-free driving stack in the industry has been authorized under a condition that assumes it might fail at any moment.
Now look east, where a different regulatory ladder is being climbed.
WeRide's founder and CEO Tony Han shared his experience with WRD 3.0 during a two-hour livestream through Guangzhou's downtown rush hour, and the number the company put on it was "0 Intervention." Two hours. Rush hour. Downtown. The company says nobody touched the wheel.
That is a marketing artifact, not an audited safety case — a livestream is not a regulator's log. But it is a marketing artifact aimed at a fundamentally harder claim than anything in the London press release. Wayve is advertising that it got a licence. WeRide is advertising that the human did nothing.
Han framed it as a mission statement: "Intelligent driving should not only be for the privileged few." Set the slogan aside and the operational fact remains — WeRide is deploying supervised-but-live autonomy that it is willing to show running hands-off for two hours straight.
Meanwhile WeRide keeps landing metal abroad. A fresh fleet of its Robobuses just arrived in Singapore, joining the GXR and a Robosweeper already working in Punggol. Singapore's LTA runs its own autonomy regime, and WeRide is stacking vehicle types inside it.
So the two companies are climbing different ladders in different jurisdictions, and the rungs are not the same height. A London PHV licence with a mandatory safety driver is a lower rung than a Guangzhou zero-intervention run — even if the London licence generates the bigger Western headline.
This is the real cross-border test, and it is not a technology test. It is a regulator test. The question for 2026 is not which stack demonstrated in the most cities — Wayve claims 500-plus — but which regulator lets you take the safety driver out.
That is the international mirror of the question we flagged as an open thread on 5 August: in the US, the argument is about intervention rates, about how often a human has to save the machine. Here the argument is about who signs the form that says the seat can be empty. Same threshold, different bureaucracies.
For Wayve, that form has a name. To go driver-out in London, it needs DVSA sign-off — a separate approval from a separate agency from the one that just issued the PHV licences. The triple-lock did not get Wayve any closer to that; the triple-lock is built around keeping the driver.
For WeRide, the wall is different and it is geopolitical. Its Western expansion runs into the same forces that put foreign autonomy vendors on covered lists — a dynamic worth naming and then leaving alone. The point is narrower: the company with the more aggressive zero-intervention story faces the harder path into Western regulators, while the company with the softer supervised licence has the home-field advantage in one of them.
Which brings us back to Wayve's actual argument for why the seat should someday be empty.
It is not miles. It is simulation. Wayve is leaning on GAIA-4, its world model, to make the safety case — running "counter-factual replay" of cyclist and pedestrian interactions in closed loop, generating the edge cases rather than waiting to drive into them.
The bet is explicit: that a world model can substitute for millions of validation miles when you go to a regulator and ask for driver-out permission. That you can prove safety in simulation faster than you can accumulate it on the road.
That is a genuinely contrarian position, and it is the one that matters. If GAIA-4 convinces DVSA, the no-maps pitch finally cashes out into the thing the headlines already claimed. If it doesn't, Wayve stays a very sophisticated minicab with a person legally in charge.
So watch what gets published, and watch the verb. "Granted a licence" is doing enormous work in the London coverage; it means a human is still responsible. "0 Intervention" is doing different work in Guangzhou; it means the human, for two hours, was not.
The 2026 milestone that counts is neither of those. It is the first sentence from a regulator — DVSA, LTA, or a Chinese authority — that permits the seat to be empty for paying passengers, with the safety case attached in public.
Until someone publishes that, the headline verb and the regulatory reality remain what they were on 5 August: a step closer, with a driver still behind the wheel.