← All mobility signals
Governance

The Careful Human Standard

Regulators are writing the safety bar for driving software as a careful, competent human, and no two prove it the same way.

What is it?

The safety bar for driving software is being written as a person: a careful, competent human at the wheel. The sentence travels well; the proof does not. Each regulator decides for itself how a trained model shows it drives like that person, and no shared number exists yet.

Why now?

Britain opened its draft safety principles for consultation in June 2026, and China approved GB 44721-2026 in July. A US notice on exemption guidance followed within days. For the first time the benchmark is written down in several major markets at once, each with its own method.

Who’s building it?

Transport ministries and standards bodies write the bar; makers, test houses and simulation firms supply proof.

Transport ministriesStandards bodiesDriverless-car makersTest housesSimulation firms
The full read
Why now, who is building it, where the market goes, and what stays unsettled.
Get weekly →
01 · The market
Who’s in it
Builders — makers of driving systems assembling evidence for each market.
Buyers — riders, fleet operators and road users asked to trust a careful-driver claim.
Gatekeepers — transport ministries, standards bodies and type-approval authorities in each market.
What’s at stake
One driving model must pass behavioural review, track tests and exemption terms separately, so proof costs multiply by market.
Whether a careful human can be measured at all.
Public trust in a benchmark without a figure.
What’s unsettled
Whether any regulator will set a numeric collision target for driverless cars.
How a behavioural principle is tested against a trained model.
Whether evidence accepted in one market will count as proof anywhere else.
02 · How the demand matureda progression, not a timeline
Novelty
Regulators borrow the careful driver as a benchmark the public already understands.
Retention
The same phrase settles into British, Chinese and draft rules as common language.
Willingness to pay
Makers pay for third-party tests, simulation and auditable fleet data in each market.
Category
Proving human-level driving becomes a standing service between makers and regulators.
03 · Where the market goesthe forks, not a prediction
→Convergence on proof: markets agree a common test suite, and one dossier travels with the driving model across borders.
→Each market keeps its own method, and makers build separate evidence for every approval.
→Fleet data grows large enough that regulators swap behavioural review for collision rates against humans.
→A serious crash forces one market to harden its benchmark into a number.
Where this lands turns on whether regulators can agree what counts as evidence that a trained model drives like a careful person, not on the phrase itself.
04 · The Driverless Deployment deskadjacent market, moving now
Miles Without a Driver
Robotaxi fleets grow before shared benchmarks exist.
Waymo runs an estimated 3,000-vehicle fleet nationwide.
SF Standard
Certified Before Release
Independent crash testing extends to AI tools themselves.
A child safety lab launches independent crash testing for AI.
CNN Business
Trucks Without Drivers, By Lane
Driverless freight wins a major commercial customer.
McLane will deploy driverless big rigs with Aurora across the Sun Belt.
CNBC
05 · The Wire
01
Britain proposes a careful-human minimum, no numeric target.
GOV.UK (Department for Transport) ↗
02
China’s L3/L4 standard benchmarks a competent, attentive driver.
CnEVPost ↗
03
NHTSA asks for comment on interim guidance for AV exemptions.
Federal Register (NHTSA) ↗
04
Dutch regulator audited a maker’s statistics for a driver-assistance approval
RDW ↗
05
UN forum adopts global driverless rules built on a competent-driver benchmark.
UNECE via Connected Automated Driving ↗
The panel on this signal
They agreed on
–Regulators anchor on a careful human because a trained driving model cannot be judged by reading its code.
–The benchmark is shared; the proof is not.
–Britain’s text is a consultation and the US item a notice.
The road-safety statistician disagreed on
–Careful drivers have no agreed crash rate by road type, so proving a match needs vast fleet mileage, not principles alone.
Know more about the method →
✦How this signal was built
01 · Panel
A Delphi panel of 4 AI analysts, each reasoning independently.
→
02 · Debate
They challenge and revise each other’s reads over two rounds.
→
03 · Evidence
Every claim is tied to a cited, linked source on The Wire.
→
04 · Synthesis
A moderator writes it into one read, disagreement and all.