The development

Uber and Wayve launched supervised autonomous rides in London in September 2026. A trained, TfL-licensed private-hire driver remains onboard during the initial phase. That qualification is important: this is a real operating development, but it is not yet a settled driverless insurance model.

What changes for insurance

Conventional motor pricing centres on a human driver, a vehicle, use and location. Autonomy changes the control structure. Driving control, operational control, system responsibility and evidence may sit with different parties.

The exposure therefore becomes more than driver behaviour. It may include the automated system, its software version, operating area, handover logic, sensor condition, fleet maintenance and access to time-stamped incident data.

The working hypothesis

Some driver-caused collisions may reduce. Other exposures can grow: repair complexity, technology dependency, correlated software defects, service interruption and disputes where the system state is unclear. Pricing therefore begins to move from an individual driver towards a risk system comprising vehicle platform, automated driving system, operator and operating domain.

THE EIGHT TESTS

A lower headline frequency is not the whole result.

01

Potentially lower, but not uniformly

Frequency

Automation should remove some collisions caused by distraction, fatigue and poor judgement. It also introduces failures in perception, handover and operation outside the system’s intended domain. The first pricing mistake would be to apply one blanket frequency reduction across every road, weather condition and driving mode.

Pricing implicationMeasure claims per automated mile and per human-driven mile, split by operating domain and system version.
02

Likely upward pressure on repair cost

Severity

Fewer accidents do not automatically mean cheaper claims. Sensors, calibration, specialist labour, software diagnostics and longer repair cycles can raise damage and mobility costs. A low-frequency book can still underperform if each loss becomes materially more expensive.

Pricing implicationTrack repair cost, calibration, downtime and replacement-vehicle days—not only accident counts.
03

Structural change required

Exposure basis

Driver age, experience and annual mileage become incomplete descriptions of exposure when control moves between a person and an automated system. The denominator needs to show who—or what—was driving, where, in which conditions and on which software release.

Pricing implicationBuild exposure around miles by mode, operating domain, vehicle platform, software version and operator.
04

New correlation pathways

Tail and accumulation risk

A defect, faulty update, mapping issue, cyber event or shared supplier failure may affect many vehicles at once. Risk that looked geographically diversified can become technologically concentrated, creating a different form of accumulation.

Pricing implicationMap common software, sensors, cloud services and fleet dependencies across the insured portfolio.
05

High selection risk while evidence is scarce

Price response and selection

Early prices will be built from limited and uneven evidence. Operators with stronger systems and cleaner data will know more about their own risk than insurers do. A simple autonomy discount could attract the risks for which the discount is least justified.

Pricing implicationUse controlled pilots, data-access conditions and narrow rating cells before making broad price promises.
06

Risk of pricing before credible experience

Capital vulnerability

Frequency optimism can obscure repair inflation, correlated loss, product liability disputes and slow evidence resolution. New capacity may price the technology story before it understands the complete risk system.

Pricing implicationSet explicit uncertainty margins, accumulation limits and review triggers while credible experience develops.
07

Seamless protection, regardless of control mode

Customer expectation

A passenger or vehicle owner will not expect an insurance gap because responsibility moved between driver, operator, manufacturer and software provider. They will expect clear cover, rapid mobility support and a claims process that can obtain the right evidence without making them arbitrate the technology chain.

Pricing implicationDesign the customer journey around continuous protection and insurer-led evidence gathering.
08

Test foreseeable harm before scale

Duty to act

Poorly explained handover rules, inaccessible event data or cover gaps can produce foreseeable harm. Firms should test whether product wording, pricing, support and claims handling still deliver good outcomes as control shifts between human and system.

Pricing implicationMonitor declined claims, disputes, delays and outcomes by driving mode; intervene where the design produces harm.

THE REQUIRED CONCLUSION

Test now. Do not price the headline.

The near-term case is neither “autonomy makes motor safer” nor “technology makes losses worse.” It is a structural redistribution of risk with insufficient mature experience. The defensible response is a bounded test with better exposure data, explicit uncertainty and accumulation controls.

Direction
MixedLower human-error frequency; higher severity and technology dependency.
Magnitude
StructuralThe exposure unit, evidence chain and liability map all change.
Timing
EmergingSupervised operation is observable now; mature loss experience is not.
Evidence
Observed + inferredThe operating development is observed; most insurance impacts remain hypotheses to test.
Response
Test nowPilot narrowly, secure data, cap accumulation and define decision triggers.

MINIMUM VIABLE EVIDENCE

What I would measure first

  • Automated and human-driven miles, separated by operating domain and system version.
  • Collision, near-miss and intervention rates by mode, road type, weather and time.
  • Repair cost, calibration cost, downtime and replacement-vehicle duration.
  • Handover state, event-data completeness and time needed to establish responsibility.
  • Shared software, sensor, cloud and supplier dependencies across the portfolio.
  • Claims disputes, declines, delays and customer outcomes by driving mode.

Follow The Exponential Insurance Brief for future applications of the Lens to emerging technology and insurance economics.

Subscribe free

Primary sources

Disclosure: The factual summary is drawn from the named sources. The insurance conclusions are my interpretation and should be tested against emerging claims and exposure data.

← Back to the Pricing Actuary’s Lens