Autonomous vehicles are not just testing London’s roads, they are testing the foundations of motor insurance. As autonomy deepens, liability is migrating from the driver to the product.
“Liability will sit mainly with the technology provider, as autonomous systems replace human decision-making,” said Simon Smith, claims director at Carpenters Group. “As control becomes fully automated, responsibility moves to the product, software, and telemetry, under strict UK safety regulation.”
That shift reframes the risk. During UK testing, reports of system disengagements, where control passes from AI back to a human supervisor, highlight the operational reality of shared control. But as full automation approaches, the transitional grey area narrows. If decision-making is embedded in code, liability follows the code.
For insurers, that is not a technical nuance. It is a structural reallocation of responsibility.
The UK’s road environment adds further complexity. High pedestrian density, heavy cyclist traffic and persistent roadworks create a testing ground that differs materially from more controlled US deployments.
“The UK’s busy urban roads, especially areas with high pedestrian and cyclist activity, create additional uncertainty for autonomous systems, which may make some insurers more cautious,” Smith said. “As far as underwriting appetite is concerned it will be interesting to see how the industry responds to an autonomous system.”
Caution here is not simply about frequency or severity. It is about classification. Autonomous vehicles do not sit cleanly within traditional motor lines, yet nor are they purely product liability.
“We are moving into a more comprehensive suite of insurance requirements and possibly a layered offering, due to migration of traditional underwriting risk management, into data services and telemetry architecture, creating the new risk model,” Smith said.
That layered approach reflects the dual exposure at play. A robotaxi operating on UK roads will still require core liability protection aligned to its physical presence in traffic.
“The robotaxi will require a core liability insurance proposition, which would cover the changing technology and telemetry demands for any robotaxi operated in the UK,” Smith said.
At the same time, the exposure increasingly resembles that of a technology provider.
“These types of autonomous product are already being offered in the USA, such as the errors & omissions insurance,” Smith said.
In the US market, that form of cover is designed to respond to technical mistakes and erroneous system behaviour, including failures that could lead to third-party bodily injury, property damage or operational interruption. It reflects an acceptance that as autonomy deepens, claims may originate not from driver error but from software architecture.
Traditional underwriting considerations do not disappear. Motor liability where a vehicle is struck by a third party remains relevant. So do fire, theft and vandalism exposures. The high-value equipment embedded within autonomous vehicles introduces significant asset risk, while cyber liability becomes intrinsic to the operating model.
The result is not the disappearance of motor insurance, but its transformation. As underwriting risk management migrates into data services and telemetry architecture, insurers are being asked to construct products capable of absorbing both road risk and software failure within a single framework.
For a market long structured around human error, that represents a fundamental shift. The question is no longer simply whether autonomous vehicles are safer. It is whether the insurance market is structurally ready for risk that increasingly sits in code rather than behind the wheel.
