A count of takeovers would mislead, and London's rules for these cars are tougher than a minicab's

This paper wants a number. How often does the driver in a Wayve car take the wheel? It treats the absence of that number as the story. I think the number would tell you less than it seems to, and that what Transport for London has built deserves more credit than it is getting here.

Start with the number. A takeover is a decision by a person. A driver told that safety comes first will take the wheel whenever in doubt, including when the software would have coped. Count those moments and you are measuring caution as much as competence. Two fleets running identical software could post very different rates because one trains its drivers to step in early and the other does not. A single figure, published without that context, would be read as a safety score. It is not one.

Now look at what TfL did require. A private hire operator normally has 48 hours to report a safety incident. Wayve has two hours to pick up the phone for a serious one. It must send a first report within a day and a full one within a week, summarise every incident each fortnight and every near miss each month, and say each time whether a person or the software was driving. That last field is the one that matters, and TfL asked for it. The rules hold whether or not a passenger is aboard.

The severity scale borrowed from Wayve sounds sinister. It is also the obvious way to get consistent reports from a company that already grades its own incidents, rather than making it run a second scale alongside the first.

And the refusal to say whether reports exist? Regulators get early, candid information when not every message turns into a headline. TfL said so in plain terms. You may not like it. It is not unusual.

Last, the paper quotes Wayve's 2022 call for a regulator obliged to publish safety findings. That was about the full regime for cars with nobody at the wheel, which has not started. This is a supervised trial with a licensed person in every car. Holding it to the standard for an empty driver's seat is the wrong test.

The house will call this a secret. A better word is a trial, run with more reporting than the minicab beside it.