Your insurer may cancel you ninety seconds before the crash

A product concept circulating among motor insurance actuaries would retire the annual policy. Instead of a premium set once and held for twelve months, the contract would be repriced continuously from telematics data — speed, braking, road surface, hour, the behaviour of the car in front — so that the customer pays for the risk being run rather than the risk expected. In the version now being modelled, the price would be allowed to move far enough that cover briefly lapses.

The mechanism would not be a refusal. Nobody would be declined, nobody would be dropped, and no letter would be sent. The policy would simply become, for a defined number of seconds, more expensive than the standing authorisation the customer had given — and would go dormant until the number came back down. Ninety seconds later the road would be dry, the price would fall, and the policy would resume, having been unavailable only for the interval that mattered.

“Nothing is cancelled,” a spokesperson for the working group would say. “The policy is offered continuously at a price that reflects continuous risk. The customer remains free to accept it at any moment.” Asked how a driver would accept a price while aquaplaning, the spokesperson would describe the question as an interface problem and refer it to a different department.

The legal difficulty would arrive with the definition of the insured event. Claims law assumes an event has a moment, and that cover either existed at that moment or did not. A contract that changes state faster than a car can skid would turn coverage into a question of timestamps, settled between two clocks that belong to the same company.

Regulators would be expected to ask the obvious thing first: whether a policy that is priced out of existence in the seconds before an impact is insurance at all, or a subscription to a risk assessment. Consumer bodies would raise a smaller and more awkward question — whether a driver can be told, in the moment, that they are currently uninsured, and whether anybody wants that notification to arrive at that speed.

The customer-facing version would be gentler. The app would display a coverage confidence indicator rather than a price, a colour rather than a number, and would suggest a change of route in the same tone it uses for traffic. Internal testing would find that drivers tolerate the amber state well and tolerate an explanation of the amber state very badly.

No such product has been filed anywhere. But every component is already sold separately: the sensors are in the car, the pricing models are in use, the driving data has a market, and the industry is moving from paying for losses towards paying on measurements. The only missing piece is the nerve to run the calculation in real time, and nerve is historically the cheapest input in the business.