How Insurance Regulators Are Responding to Driverless Fleets

Insuring a fleet with no drivers in it is a different actuarial problem than insuring a fleet of personally-owned cars, and regulators are still working out exactly how different.

Traditional auto insurance prices risk around the individual behind the wheel: driving history, age, claims record, the whole profile an underwriter builds around one person. A driverless fleet removes that person from the equation entirely. There's nobody to assess. So commercial fleet insurance for these vehicles gets underwritten against the operating company itself, its safety record, its vehicle fleet, its operational history, rather than against any individual driver, because there isn't one to price against.

That's a genuinely different underwriting model from anything the broader insurance industry deals with in taxi or rideshare contexts, where a human driver is still somewhere in the risk equation even under a commercial policy. Take the driver out entirely and the underwriter is left pricing the software, the sensor suite, the maintenance program and the company's operating history instead, an evaluation that looks closer to industrial or product-liability underwriting than personal auto insurance ever did. Regulators writing new AV insurance rules are, in effect, asking underwriters to build that evaluation largely from scratch rather than adapt an existing model.

The absence of a driver also removes a specific complication that has shaped conventional auto insurance in the US for decades: the split between no-fault and at-fault liability regimes that different states apply to ordinary drivers. That split exists to sort out competing claims between two human drivers who might each bear some share of responsibility for a collision. A driverless fleet vehicle has no driver to assign partial fault to in the first place, which doesn't make the underlying legal question disappear so much as redirect it entirely toward the operating company, the software, and whatever else in the chain might have contributed to a failure.

Some regulators have responded by writing AV-specific insurance requirements directly into their deployment-permit frameworks: minimum coverage levels, specific product requirements, conditions attached to the permit itself rather than left to whatever the market happens to offer. Others haven't gotten there yet, and the requirements that do exist vary substantially from one state or country to the next. None of it is settled the way conventional auto insurance rules are settled after a century of precedent.

There's an underwriting advantage buried in this shift that's easy to miss. A driverless fleet generates continuous, detailed operational data, exactly how each vehicle behaved in a given situation, rather than the indirect proxies, self-reported mileage, credit-based scoring, a multi-year claims history, that conventional personal auto insurance has long relied on to estimate an individual driver's risk. An underwriter pricing a fleet against its own operating history is, in principle, working from a far richer and more direct data set than one estimating an individual driver's future behavior from a rearview mirror of past claims. Whether insurers and regulators have actually built the tools to use that data well yet is a separate question from whether the data itself exists.

That variation, and the pace at which it's still changing, is exactly why anyone making a decision that actually depends on the specifics, coverage minimums in a particular state, what counts as an adequate product for a given permit, should go to the relevant regulator directly rather than treat this as the final word. A closer look at the open liability and insurance questions covers where the harder disputes are still unresolved.