Insuring Autonomous Vehicle Risks
Written in collaboration between
ibott & IOA
Insuring Autonomous Vehicles is difficult. The technology is new and the data linked to insurance is sparse. Currently the marketplace is small though and the insurance requirements are largely focused on companies performing testing, often with little if any interaction with third parties. The ultimate opportunity for insurers though is unquestionably huge! Today, most reports will show Auto premiums at circa 50-55% of all non-life insurance premiums and if the future is going to shift towards automation, the entire insurance product linked to automobiles will need to change. However, creating this insurance product is not without its challenges.
The Coverage
Focusing on insuring companies testing and operating in the United States, insurance is required to register the vehicles and subsequently primary auto liability insurance is typically purchased on an admitted paper basis. Most individual States then require AV companies to purchase excess insurance (limits vary by State but most adopt a minimum requirement of $5,000,000). Many partner contracts require even higher limits. On an admitted coverage basis, insurers cannot exercise flexibility in the breadth of coverage they provide and the rates they apply. It forces insurers to cover these new and emerging exposures with a traditional auto product, which many will agree is unfit for purpose. Many AV companies are surprised to learn that their insurance product is the same auto policy that has been in the marketplace for years. Indeed, there is no mandate that a human drive the vehicle. There’s never been an alternate option until now.
The exposures to the vehicle of yesterday were simple: physical damage and liability. The Autonomous Vehicle of tomorrow adds exposures of:
perception hardware failure
software failure
network intrusion.
Those are all covered by different insurance products. The claims difficulties begin with a forensic analysis to determine what caused the accident (not easy, see the Uber in AZ case). Once a cause is determined, wrangling amongst different carriers insuring auto, product liability, and technology errors & omissions, and cyber begin. This lack of clarity is of course a result of insurance placements often being disjointed, with coverage often purchased for different lines separately, with sometimes different intermediaries involved. As a result, there can often be coverage for specific perils potentially under multiple policies.
For such a large new and emerging risk the best fit would be a new line of insurance to adequately address this specific risk. Will we soon see the first Autonomous Vehicle Liability policy? If so, what will it look like?
Vehicle Ownership
When (not if) autonomous vehicles become part of our present and not of our future, one element that will be interesting is the effect on vehicle ownership. Will individual consumers replace their traditional automobiles with an autonomous vehicle? Industry consensus to date has suspected that we will not see a large amount of private ownership of AVs. The average owned car sits idle for over 95% of the year. In an autonomous vehicle future, we envisage that there would be a shared AV operating system as part of a wider mobility solution. Permitting that shared solution as ample supply to meet demand and operates nationwide, why would someone necessarily need to own an AV and would it be economically viable? As with everything else in our lives, the worldwide COVID-19 pandemic has created new challenges. We now have to consider how the public and various health organizations and governing bodies will view further adoption of shared travel. How will sanitation change?
How does ownership impact insurance? Beyond insurable interest, it clearly adjusts which product responds. Imagine a future fleet risk of autonomous vehicles. The owner is an investment fund. The “driver” is AV software. The mapping and user interface is run by a TNC. Chances are, unless the vehicle has an accident as the result of some deferred maintenance the responsibility will not rest with the ‘owner’. It will fall to the other parties involved as tier 1 suppliers, partners, and service providers.
Ownership highly impacts what an insurance solution looks like. If we are looking at a shared AV model then we are considering commercial lines rather than personal lines. We are also looking at a few sophisticated buyers rather than a large pool of buyers. It is likely that these sophisticated buyers will want to share in risk, either via taking a large/retention or deductible or going even further and establishing a captive and looking at a more comprehensive self-insured solution.
How do you rate Autonomous Vehicle Risks?
Traditional auto insurance rating factors have largely focused on who is driving, what are they driving and where they are driving. The evolution of telematics in the Auto insurance industry has allowed insures to start rating on how they are driving, but that is still in its infancy. The removal of a driver removes one of the key rating factor categories, but it also significantly amends some of the other key rating factors – what is being driven (the vehicle itself) and where is it being driven.
However, it is not likely that insurers will now adopt a very simplistic rating model. In fact, with additional data being made available, the rating may become more complex. The rating factors rather than being removed, will simply change to consider different risk metrics. Insurers will start discussing:
hat makes a good AV risk versus a Bad AV risk?
Insurers will need to develop new rating models for this new and emerging risks. Can they use and rely on the AV disengagement statistics that most AV testing companies submit to the various different State departments of transport? Will they utilize risk engineers to evaluate the different software and capabilities of the individual risks? What exposure metric will be used? Will insurers adopt the traditional per vehicle rating or apply a usage based metric such as per mile? Any solution should really be created with collaboration with the AV testing companies themselves of course.
The industry is beginning to establish safety standards, as evidenced by (UL) 4600 (released in 2019 by Underwriting Laboratories and Edge Case Research) and NSTHAs recent AV Safety Principles. The hope is that these minimum standards will fill some of the vacuum of consistent and cohesive regulatory standards and define areas of responsibility in this ground-breaking industry. The approach taken is to provide “risk analysis and safety-relevant aspects of design process, testing, tool qualification, autonomy validation, data integrity and human-machine interaction for non-drivers”
There is a very real risk that legacy insurance carriers fall behind emerging Insuretech companies in their willingness to invest in innovation and data analytics. This situation may be compounded by state regulations on which criteria can be included as rating factors. Currently, California does not allow telematics data on driver habits (near misses, hard braking, etc.) in its rating. What sort of challenges will the industry face in getting regulators up to speed on the unique promise of AVs?
Our advice to AV companies when approaching insurance purchases
Our primary suggestion would be invest in finding the right partners. Work with insurance professionals who are focused and committed to this space. Collaborate with those professionals (both intermediaries and carriers) to create a fit for purpose insurance solution. The alternative is to allow others to create the solution in their image and then living with the consequences to your business of the result.
