Brian Wong and Anne Buhl Bjelke (of Bech Bruun) consider what the regulators and governments are doing, along with the key current and future issues, for driverless cars and motor insurance in Europe.
This article first appeared in the September 2017 issue of the Committee Update of the Insurance Committee of the Legal Practice Division of the International Bar Association (Vol 23, No 2), and is reproduced by kind permission of the International Bar Association, London, UK. © International Bar Association.
Since the article, in the UK the Automated and Electric Vehicles Bill has been introduced into Parliament. See our separate Legal Update.
The potential for driverless vehicles to transform the motor industry and motor insurance industry still looms large. By some studies, over 90% of road traffic accidents continue to be caused by driver error. This is notwithstanding great strides in autonomous advancement such as Autonomous Emergency Braking (AEB), anti-lock brakes, automated parking and sensors, adaptive cruise control and lane assistance/warning systems – all of which are now considered mainstream vehicle assistance features and are factored into insurance groups. AEB alone is estimated in Europe and Australasia to reduce real-world rear-end crashes by 38%.
With the development, adoption and deployment of driverless vehicles now fast entering the mainstream, what do insurers need to be aware of and anticipate?
What issues do the insuring of driverless vehicles raise?
The arrival of true driverless vehicles (or 'connected and autonomous vehicles' or CAVs) onto roads poses some rather unique challenges (and opportunities) for the motor insurance industry in terms of pricing and grouping risk, writing new insurance products, liability scenarios and access to vehicle data.
Much of this relates to the transfer of driving function away from a human driver to a vehicle and driving technologies and the consequent change in dynamics of the insurer-insured relationship (in the UK, motor insurance insures the driver and not the vehicle itself). This process of transfer is likely to be a gradual one in most places and SAE International has described a 6 level taxonomy of driving automation from full driver control with no autonomous assistance through variations of shared control to full system control. Each level of automation potentially offers its own insurance challenges.
Partial Automation Driver assistance system undertakes steering and acceleration/deceleration using information about the driving environment with the expectation that the human driver will perform all other driving tasks.
The liability framework
In a Level 5 fully connected and autonomous car, liability in practice moves entirely away from the driver. This fact is illustrated most starkly through those developing Level 5 technology today where the vehicles simply have no steering wheel at all. An accident caused by the car at this level could be equated to a 'system error'. That is not to say that liability issues are straightforward of course – ultimately even a system error needs to be diagnosed and a driverless vehicle system is a complex system comprising vehicle, software, local and central control systems, highway systems and connectivity infrastructure for which many different stakeholders may be responsible. In turn, that may take on an added degree of complexity when parts of the system become powered by increasingly autonomous artificial intelligence systems making dynamic decisions in real-time.
At levels of autonomy below Level 5, the driver remains an additional factor in the liability framework and, crucially, it becomes important to understand who or what has mechanical control of the vehicle at any particular point and how the human-machine interface deals with handover of driving functions from machine to human and vice versa. VENTURER, one of the CAV pilot projects in the UK in which Burges Salmon is participating (together with AXA Insurance), has been looking into this crucial aspect of handover.
In all cases, whilst fully autonomous and/or highly autonomous vehicles share public road space with vehicles at Level 2 or below, there remain potential liability confusion and issues in such an ‘open system’ where machine drivers will at all times be interacting with human drivers. This would contrast with a closed autonomous vehicle system where all vehicles on roads are connected and autonomous.
Standardisation and group ratings
Understanding and pricing the market is key for insurers to be able to bring products to the market to meet demand and facilitate the adoption of this new technology. To do that there needs to be a degree of standardisation and regulation in terminology and technologies across the motor industry to ensure integrity across classes of connected and autonomous vehicles. This is critical not only so that customers can know what they are buying and driving but to allow driverless vehicle makes and models to be grouped as traditional vehicles are for insurance purposes. There has, for example, already been comment on the difficulty in categorising or comparing various models of Tesla for insurance purposes.
Standardisation and regulation may also be required for elements of supporting vehicle connectivity and operating infrastructure. This would include the networks used to connect and guide vehicles both physical (e.g. on-street connectivity equipment) and virtual (e.g. the geo-location / mapping capability by which driverless vehicles navigate).
Standardisation applies equally to the vast amounts of data that will be collected and generated by driverless cars. How that data is validated, protected and shared is key to the success of CAVs. Ultimately, a driverless car system is fuelled by data to allow it to work. Another pilot project in the UK, FLOURISH, in which Burges Salmon is participating (again together with AXA Insurance) is looking at (amongst other things) the critical issue of cyber security and wireless communications. Burges Salmon and AXA are specifically focussed on the legal and risk environment which supports the transformative technologies of CAVs from both a data and cyber security perspective.
The CAV ecosystem involves many stakeholders - from software developers, vehicle manufacturers to insurers and local authorities. It is envisaged that only some of the data generated by CAVs will need to be shared for stakeholder purposes but there should be a degree of standardisation around those. For example, accident data will be critical to understanding causes of accidents for liability purposes. Robust contemporaneous data could present a step change in the way that claims are dealt with and the efficiency in which they are resolved benefiting both insurers and insureds.
What approaches are industry, government and regulators developing?
Recognising the steps already being taken towards ever greater autonomy and the potential safety benefits of the technology as well as social, economic and efficiency factors, there is a significant and growing base of support across industry, government and regulators. Recent years have seen an acceleration in efforts to facilitate, develop, test and bring the technology to market. If anything, the main stakeholder remaining to be brought on board is the travelling public itself. Many local and national surveys attest to the current levels of reluctance of drivers and passengers alike to trust driverless vehicle technology. Winning over public confidence and acceptance is critical to the successful uptake and growth of CAV technology.
Steps towards new legislation have been or are being taken in various member states in the EU where governments are undertaking legal and regulatory changes to permit the testing and eventual mass marked introduction of driverless vehicles on roads. In January 2016 the EU Commission launched a new High Level Group for the automotive industry named "GEAR 2030". The purpose of GEAR 2030 is to assist the Commission in developing recommendations to reinforce the competitiveness of the European automotive value chain and to establish work plans for a roll out of driverless vehicles in the EU up to a current time horizon of 2030.
In April 2016, the transport ministers within the EU signed the Amsterdam Declaration on cooperation in the field of connected and automated driving. The declaration aims inter alia to encourage member states to identify regulatory barriers with regards to testing CAVs in the EU, and to co-operate and work towards developing a coherent European framework for the deployment of interoperable CAVs. This was followed in November 2016 by the Commission's "A European strategy on Cooperative Intelligent Transport Systems, a milestone towards cooperative, connected and automated mobility" setting out plans for EU level legal frameworks and CAV deployment in 2018/2019.
Following the intent of the Amsterdam Declaration, for example, Denmark has taken necessary legislative steps and on 1 July 2017 an amendment to the Danish Road Traffic Act became effective. The amendment allows for the Danish Government to grant permission to test driverless vehicles on the public roads of Denmark as part of a pilot scheme. It is expected that the first such permissions will be granted in 2017.
Since around 2015, the UK Government has been committed to growing the CAV market in the UK and making the UK a world leader in the technology. It has set out a comprehensive methodical policy pathway for the testing and promotion of CAVs and CAV technology and is consulting on regulatory change to facilitate CAV adoption and to clear legal obstacles in phases as the technology develops. The UK Government has committed funding in excess of £100m to a variety of pilot and research and development projects across the country in addition to separate funding of similar orders for associated powertrain development (e.g. battery development) and low emissions technology.
From a legal and regulatory position, it is plain therefore that governments are rapidly gearing up for CAV technology.
Insurers have recognised from the outset that they too have a role in facilitating the development and adoption of this new market, together with lawyers, government and regulators, by smoothing out the uncertainties in the liability model in particular.
Regulatory and insurance uncertainty can lead to patchy, uneven and unsatisfactory development as well as unilateral stakeholder action. The rapid roll-out of near 'driverless' technologies which require, as a condition of use, full driver attention and contact throughout operation, over-the-air autonomous fundamental feature updates and pronouncements of manufacturers such as Volvo that they will be responsible for accidents involving their cars in autonomous mode are examples of what happens when regulation and insurance potentially sets up a conflict between the legal liability model and how the current technology is being used in practice.
In the UK, 11 insurers led by the Association of British Insurers (ABI) and Thatcham Research have come together under the banner of the Automated Driving Insurance Group (ADIG) to consider key issues relating to automated driving on UK roads, particularly concerning insurance and liability. ADIG will feed into industry policy and work with the UK Government on shaping the future of automated vehicle use in the UK and, where appropriate, in the EU and internationally. The first UK insurance products for driverless cars are already being promoted.
In the UK the current position is that third party motor insurance is compulsory and, if any autonomous assistance feature failures are causative, that may affect the driver's liability position and potentially lead to product liability claims made against manufacturers. This position is maintained through the current CAV testing phases where the UK requires that a driver is present at all times ready to assume control of any test vehicle and is insured on that basis.
The UK Government has consulted widely on the insurance model going forward and has previously published draft legislation in the form of the Vehicle Technology and Aviation Bill (due to be refreshed in 2017) which would have extended the compulsory motor insurance regime to include product liability aspects so that both drivers and vehicles are covered under the same policy at automation levels equivalent to around Level 4 or 5.
This 'single insurer' model is supported by the UK insurance industry and would mean that the public could be reassured that in any accident involving a highly autonomous or fully autonomous vehicle there would be a straightforward single point of recovery regardless of whether or not the driver of the driverless car or some system component was at fault. Instead such liability issues would be secondary issues dealt with 'under the surface' between insurers and other CAV stakeholders. At Levels 3 and below, the status quo would prevail and the driver alone would be the primary insured. Where the driver was not at fault for the accident but the CAV technology, product liability issues would arise.
The preparatory works of the Danish amendment of the Danish Road Traffic Act shows that both the civil and the criminal liability regime were considered closely. In respect of civil liability, the Danish Act maintained the current regime for liability and duty to insure to a large extent. A third party motor insurance is compulsory according to Danish law. Further, the owner or user is subject to a strict liability (liability without fault).
Under the amendment of the Danish Road Traffic Act allowing for testing of CAVs neither the obligation to insure nor the rules on liability have been amended. The add-on to the current liability and insurance regime is who carries the liability and duty to insure in case of driverless vehicles. For conventional vehicles the obligation to take insurance is on the owner or user (not the driver) of the vehicle and a third party motor insurance has to be taken out by the owner or user on each individual vehicle. For driverless vehicles it is the test license holder who is liable for damages, and it is also the test license holder who is required to insure the vehicle against third party risks. The reasoning behind this is that it is the test license holder who has the opportunity to design and plan the use of vehicles in a safe way as well as the test license holder has the interest (financial as well as other interests) in mitigating risks. Thus, in case of a driverless vehicle the test license holder is subject to the exact same liability regime and insurance requirements as the owner or user of a conventional vehicle and the third party is covered in the exact same way in case of an accident involving a driverless vehicle.
Like the UK proposals, the Danish amendment of the Danish Road Traffic Act does not amend the product liability regime. Given that rules on product liability would also apply where liability involving a CAV is involved, the question is, whether the third party motor insurance insurer can claim recourse for damages paid out. Such recourse may lie against, for example, the software supplier or the manufacturer of the vehicle.
Given this critical role of product liability regimes in insuring of CAV technology, it is likely that product liability reform will also be considered by governments and the EU as part of their preparations for CAV technology and deployment.
The EU Commission has expressed its understanding that the Product Liability Directive 85/374/EEC may also apply to software. At the time of writing, the Product Liability Directive is under revision and the EU Commission is expected to clarify the position on product liability for software, which may impact the liability distribution between drivers, owners and manufacturers. This will raise questions in respect of under which circumstances a software program is to be considered defective under the product liability rules. If the software acted as it was programmed for, but the action was inappropriate or not safe to third parties, would the software be considered defective? This directly engages some of the design / ethical considerations around CAVs, and – more pragmatically – questions how standards for software and AI should be designed to balance:
- the public interest of safety of all parties in the traffic; and
- socio-economic interests in development of new technology and enterprises’ interests in bringing commercially attractive products to the market without sowing confusion in the resulting liability framework.
The key issues
Standardisation of data collection and protection standards and protocols for sharing of accident data. From a Danish perspective and according to the amendment of the Danish Road Traffic Act it is a condition precedent for a license to test driverless cars that a detailed plan regarding collection, registration, storage of and passing on of data is developed and applied. One of the purposes of the data collection and storage is that in case of an accident the data must be available for investigation. Such considerations must be taken into account and required as part of eventual CAV deployment beyond testing phases.
Development of vehicle and technology standards and regulations to support development of autonomous vehicles in a safe, coherent and co-ordinated manner.
Clarity from government, as reforms develop, as to the regulatory treatment of CAVs and CAV systems (pilot and commercial forms) and future intentions where treatment will differ from ordinary road vehicles. At a minimum, a level of standardisation and harmonisation of legal and frameworks will be desirable if, for example, some of the concerns such as those expressed about uneven regulation across US states are to be avoided.