Third-party liability insurance for owners of autonomous vehicles | In Principle

Go to content
Subscribe to newsletter
In principle newsletter subscription form

Third-party liability insurance for owners of autonomous vehicles

Autonomous cars are slowly becoming a reality, accompanied by logistical, technological and legal barriers. The Polish parliament has already made some initial legislative steps by introducing, for example, a definition of an autonomous motor vehicle and a procedure for testing such vehicles on public roads. However, there is a need for further statutory regulation of insurance products and insurance companies tasked with providing compulsory third-party liability protection in a situation of great uncertainty and legal gaps.

First problem: a business model for a new technological challenge

To fully understand the issue at hand, we must first understand the insurers’ business model. Insurance is based on an inverted business cycle, where first the insurance company generates revenue by collecting insurance premiums, only later to incur its most significant costs in the form of monetary claims paid (compensation or damages). It is inverted by comparison with, for example, the traditional form of trade, where a firm first purchases goods (generates costs) and then sells them to buyers, thus generating its basic income.

In the insurance business, it is essential to properly calculate the net premium. The total amount (collected from all insureds in a given community of risks) should match the benefits to be paid by the company, i.e. the expected loss value (also understood as a fixed loss). This value is calculated using the law of large numbers and advanced mathematical models, i.e. based on statistics (public and the insurer’s own) and historical data. The aim is to identify various risks: to determine their number, frequency of occurrence, potential size, reasons for their creation, proper selection, etc. For example, if the calculation shows that for a given risk community the expected value of loss is PLN 20 million, the total net premium charged should reflect that number.

Such calculations will only converge with the actual frequency of events and the magnitude of damages when properly scaled. There is no doubt that in the absence of such data (or when there is insufficient data or the sample is too small), there is a risk of obtaining an inappropriate amount of net premium. Both overestimation and underestimation generate undesirable effects for insurance activity.

The problem of the lack of such data is already arising because, according to definitions in the Traffic Law and the Compulsory Insurance Act, an autonomous vehicle is a motor vehicle subject to compulsory third-party insurance. This means that an insurance company cannot refuse to conclude a compulsory insurance contract even though it does not yet have the relevant statistics and historical data necessary to properly identify the risks and their extent, and thus the real value of the expected loss is unknown. This problem is particularly important as vehicle insurance is one of the two largest sub-groups of Division II insurance activity (owners’ third-party liability for land vehicles and collision cover (autocasco) for land vehicles).

Therefore, actuaries can be expected to rely on scientific studies, data for example from reinsurers, and observations from test drives on public roads (although even such test drives will also be subject to compulsory third-party liability insurance).

Second problem: lack of adequate civil regulations

The second problem arises from general civil regulations on torts, i.e. events giving rise to obligations. In this context, it is worth recalling the definition of an autonomous vehicle, that is, a motor vehicle equipped with systems controlling the movement of the vehicle and allowing it to move without the intervention of a driver, who can take control of the vehicle at any time. Under the definition, every autonomous vehicle must have a driver, regardless of the level of automation of the vehicle. Currently, international standards specify five levels of automation (above level 0, no automation). Level 5 means the highest degree of automation: the car is completely autonomous, can move without a driver, perform the most complex manoeuvres, drive to a selected point, etc.

With such a high level of vehicle autonomy, the least problematic case seems to be a traffic accident where an autonomous motor vehicle collides with a person who consequently suffers, for example, a bodily injury or health disorder. In this case, under the applicable provisions of the Civil Code, the owner of an autonomous motor vehicle will be liable for damage on a risk basis (strict liability). Here, the level of autonomy of the vehicle will be irrelevant, as the risk of driving a vehicle powered by natural forces (electricity, fuel, etc) will determine liability. Potential reduction in the compensation or damages payable will depend solely on the extent to which the victim has contributed to the injury; nevertheless, a question arises as to whether the extent of the contribution may vary in light of the specific characteristics of the autonomous vehicle and its algorithm (such as atypical reactions to sudden crossing of a lane by a pedestrian, etc).

A collision between two motor vehicles, with at least one of them being an autonomous vehicle, will be more problematic. There the aggrieved parties may mutually demand compensation for damage suffered, but only on general principles, i.e. on the basis of fault. Undoubtedly, it will be necessary to determine whether the autonomous car was driving entirely independently (guided by an algorithm) or was controlled by the driver at the time of the collision. If the vehicle is being steered by an algorithm, it is difficult to assume that the driver could have been at fault for the accident (unless taking control at the right moment would have prevented the collision). It seems that in relation to the algorithm steering an autonomous vehicle itself, it is impossible to speak of independent will, as the algorithm only “obeys” certain programmed and coded rules. Thus, no fault in the traditional understanding of fault according to general principles should be attributed to it, which consequently results in a lack of civil liability. Another question also arises: what if “fault” in the collision is mixed, i.e. the driver was partly at fault and there was partly an error in the algorithm driving the autonomous vehicle?

These issues directly affect the scope of liability of the insurance company, as it is vicariously liable, on the same principles as the insured.

Third problem: lack of adequate regulations reflecting the nature of autonomous vehicles

Insufficient industry regulations taking into account the specifics of autonomous vehicles is undoubtedly another problem. In the case of compulsory insurance against civil liability of motor vehicle owners, the law in Poland includes special rules concerning such issues as insurance recourse, which may not prove sufficient in respect of autonomous vehicles. Thus, under current legal solutions, an insurance company has the right to seek reimbursement of compensation paid from the driver of a motor vehicle if, for example, the driver caused damage intentionally, after using alcohol or drugs, or fled the scene of the accident.

These examples seem not to address the specifics of autonomous vehicles. What if a level 5 autonomous vehicle was used, and the driver (essentially a passenger) was intoxicated but not driving the vehicle at all, and the driver’s taking control could not have objectively prevented the accident or collision? What if the vehicle fled the scene due to a technical error in the software (the algorithm driving the autonomous vehicle)? Would the condition of the driver fleeing the scene still be satisfied?

It also seems that the catalogue of events entitling the insurer to seek recourse should include, among other things, unlawful interference with the algorithm driving an autonomous vehicle. For example, if in the course of the proceedings for indemnification of loss under an insurance policy it is proved that interference in the algorithm directly caused the accident, the insurer should be entitled to seek recourse against the person who carried out the unlawful interference.


These issues are only the seed of the problem of autonomous vehicles in relation to the existing regulations on compulsory insurance. The dynamics of technological development and insurance technology mean that legislation relating to these issues must be brought up to speed to meet the challenges. It is vital for legislative solutions to be founded on sound principles and balance the interests of all stakeholders: insurers, insureds, and persons who may be injured in an accident involving an autonomous vehicle.

Mateusz Kosiorowski, Insurance practice, Wardyński & Partners