For insurers watching autonomous vehicles move from trials towards paid services, recent robotaxi incidents are less about fault and more about evidence, and whether current safety data is robust enough to support early pricing, reserving and liability assumptions. As real-world deployments expand, live incidents are beginning to test risk models built largely on controlled trial environments.
That tension is already visible in how automated vehicles are interacting with mixed traffic. Recent reporting has highlighted growing unease around automated vehicle behaviour, questioning whether safety claims drawn from testing conditions fully reflect the complexity of open-road operation, particularly where autonomous vehicles must coexist with unpredictable human drivers.
According to Ben Gardner, partner in manufacturing, technology and mobility at Shoosmiths, that unpredictability is a key reason insurers should be cautious about assuming near-term risk reduction.
“In the short term there is no guarantee that AVs are a lower-risk proposition,” Gardner said.
“For example, the technology will be constantly developing and learning, and AVs will be sharing the road network with human driven cars that may be less predictable and less able to communicate.”
Recent US incidents illustrate that challenge.
In San Francisco, a Zoox robotaxi was involved in a low-speed collision after a parked car door was opened into its path, resulting in reported injuries. The operator described the contact as unavoidable, highlighting how even minor interactions can quickly become contested claims events once autonomous systems are involved.
In Washington, D.C., a self-driving shuttle carrying officials during a public demonstration was struck by another automated vehicle attempting an illegal manoeuvre, causing damage to both vehicles. While no injuries were reported, the incident underscored the exposure automated vehicles still face from human-driven traffic, even in managed or showcase settings.
Operational disruption has also emerged as a consideration for insurers. A Waymo robotaxi was recently reported to have frozen at an intersection in Texas, remaining stationary across multiple traffic-light cycles while awaiting remote intervention, demonstrating how non-collision events can still create safety, congestion and liability implications.
Looking beyond incident frequency alone, Gardner said insurers must also consider how claims severity may evolve as the technology matures.
“In the medium to long term, there are likely to still be accidents involving AVs but the expectation is that there will be less,” he said.
“Less accidents would indicate less risk. However, AVs will be fitted with sensors, cameras and other expensive hardware compared to a conventional vehicle. The cost of repairing or replacing this in the event of an accident will be higher which potentially balances out the reduction in accidents as the value of claims may be higher.”
For insurers, this creates a more nuanced underwriting picture than headline safety statistics alone might suggest. Fewer incidents do not automatically translate into lower overall exposure if repair complexity and average claim values increase.
Despite the short-term uncertainty highlighted by recent incidents, Gardner said trial data continues to support the longer-term safety case for autonomous vehicles - provided technology maturity, public awareness and legal frameworks develop alongside deployment.
“The data flowing out of trials still suggests that AVs will be a lower-risk proposition and as the technology matures, public awareness increases and legal frameworks are developed, the long-term safety benefits should become more demonstrable,” Gardner said.
“Incidents along the way are, unfortunately, inevitable.”
For insurers, the challenge is not whether autonomous vehicles will eventually be safer, but whether the market is being asked to price that future before the evidence is fully there.
