Automated and Electric Vehicles Bill

Written evidence submitted by Professor James Davey, Professor of Commercial Law at the University of Southampton (AEVB 04)

Re: Automated & Electric Vehicles Bill 2017

FAO Mr Aidrian Baily MP and Sir Edward Leigh MP

I welcome this opportunity to contribute evidence to your Committee. The attached note is intended as a primer to consider the implications in practice of this Bill, rather than to call in to question the policy choices which motivated it. The AVs at the centre of these provisions have been carefully designed and tested over years. The legal vehicle does not, as yet, seem to have had the same rigorous approach. I hope in some small measure I can help the Committee improve the roadworthiness of this Bill.

I would be very happy to expand on any element of the attached that the Chairs believe would be of benefit to the Committee’s decision-making process in person or in the form of supplementary papers.

Written Evidence ‘Pathway to Driverless Cars: Insurance for Automated Vehicles’

Professor James Davey [1] *



The purpose of this note is to highlight significant legal and practical issues with the drafting and design of the Automated & Electric Vehicles Bill 2017. This is not meant to call into question the choice of ‘Option 2’ [1] as the preferred model, but to show difficulties in implementing the model. Some of these issues might be addressed by changes to wording in the Bill, others might require a more fundamental shift in approach.

Issue 1: Liability on the Motor Insurer


When considering how to incorporate AVs within the existing scheme it is important to understand that UK motor insurance compensation is generally limited to cases of established fault of humans. In practice, this is often simplified by various rules of thumb, but we are primarily concerned with finding fault. The driver at fault is liable and her liability insurer has agreed by contract to pay sums to meet those liabilities. In motor insurance, statutory intervention gives the innocent victim a direct right to claim from the insurer, but it still requires that the insured driver was liable because she was at fault. There are then two systems meshing together (1) ‘fault’ (this is the law of tort liability) and (2) insurance (to provide effective compensation):

1. Insured driver owes ‘duty of care’ in tort to injured cyclist;

2. Underwriter owes contractual duty to insured driver to pay claims within policy.

3. Road Traffic Act 1988- gives right to cyclist to merge steps (1) and (2) into single process. Prove insured driver was liable and claim insurance proceeds direct.

The design of Clause 2 is different and problematic. Cl 2(1) makes ‘the insurer’ strictly liable for losses, but without any associated liability on any insured person. It activates where there is a policy of motor insurance operative under cl 7(1): (‘a vehicle is "insured" if there is in force in relation to the use of the vehicle on a road or other public place in Great Britain a policy of insurance that satisfies the conditions in section 145 of the Road Traffic Act 1988’). But under the current system we would start with ‘who is at fault’ and then decide ‘which insurer will compensate’. The 2017 Bill does not fix any individual with liability and there is no obvious mechanism for determining which insurer pays. If four people (all of whom are insured to drive any vehicle under their policy) are in a car in autonomous mode running as a taxi, the statute says nothing other than the insurer shall pay. If we mean the person carried in the driver’s seat, then say so. If we mean perhaps instead the commercial provider of the vehicle (assuming a taxi service) then say so. ‘Use’ of a vehicle under the RTA 1988 is a fairly elastic concept and ill-suited to this purpose. The decision not to make someone liable for the use of the vehicle and then attach liability insurance to that liability is very peculiar to say the least. A recent discussion group of leading UK/US experts could think of no comparable situation in either legal system. [2]

Some of the potential legal and practical consequences of the decision to impose strict liability on the insurers, rather than make them a liability insurer (as in the current system), are listed below. Each of these could be developed in more detail if required in a supplementary paper or as oral evidence:

· It reasonably clear that the insurer is NOT liable as an insurance liability. It is liable in tort [cl 6(4)]. This would create novel legal problems, for example on the applicable limitation period; consequences of late or non-payment of the claim (e.g. relationship to the new duty found in the s. 28-30 Enterprise Act 2016) and so on. Insurers are normally liable in contract, but to meet liabilities incurred in tort by their insureds.

· Insurers will have to recoup this loss from an insured. There is no free lunch. But if the insured has no liability, she has no ability to challenge the claim or any settlement. Insureds rarely do this, but it is not clear to me that they should be removed from the process automatically.

· If insureds are not part of the process, it is hard to see how they have rights to any evidence collected during the process. Much is made of the data capture ability of AV vehicles involved in accidents. The insured may have a very real interest in knowing what happened and why.

· None of this will replace the need for vehicle manufacturers to have product liability and product liability policies. Injured cyclists, passengers and others can still potentially sue them. Suggestions that ‘no one would’ depend on the political climate. I can see many reasons why a major motor manufacturer might settle a claim for reputational reasons where an insurer might resist the claim. And vice versa. This Bill does not avoid the duplication of legal processes. This may be inevitable. But the efficiency arguments look less self-evident than claimed.

Issue 2: ‘Accidents Caused by an AV’


The notion of cause is a slippery one. In the AV environment, it is necessary to determine whether we are comparing the AV with the average human driver; or ‘reasonable’ human driver (not the same thing) or the average AV system when deciding whether the decisions made caused the accident. Even in non-fault systems, the highest courts have tended to look for some action which is out of the ordinary in order to identify a causative moment.

One example: AV systems might use LIDAR or RADAR or a combination (etc.) to identify objects in the real world. If an AV is involved in an accident and it did not see the cyclist because (e.g.) it was snowing heavily [3] , then do we compare the vehicle with the best AV on the market; the best AV in that price bracket or what a human would have done?

This is crucial for clause 2. There may be a deliberate decision here to leave this to the judges to work out. I think it should be considered in advance, especially if we wish to build consumer confidence.

Issue 3: Updates, Vehicle Maintenance and the RTA 1988


The draft Bill [cl 4] permits insurers to deny liability where certain types of preparation are not carried out; notably operating system updates. But a recent Court of Appeal judgment (UK Insurance Ltd v Holden [2017] 3 WLR 450) treated maintenance of the vehicle as ‘use’ of the vehicle, so that the motor insurer was liable under the motor insurance contract (in light of the compulsory cover required by RTA 1988).

Imagine a vehicle which is capable of human-controlled and autonomous driving. As I understand the combined effect of Holden (on the RTA 1988) and the new Bill, this means a motor insurer could exclude liability under the Bill for ‘autonomous’ liability, but might be fixed with liability under the RTA 1988 because updating software is maintenance of the car equal to ‘use’, which then leads to a loss. [4] It is not clear to me whether that this is deliberate. If it is the same insurer at both times, then we have to resolve the conflict of statutes. This is not impossible, but it is not evident that the drafters of the Bill have recognised this potential issue. If it has been noted, I suggest that it needs to be much more clearly stated as a policy choice, if the courts are to follow the reasoning.



The AVs at the centre of these provisions have been carefully designed and tested over years. The legal vehicle does not seem to have had the same rigorous approach. This is not primarily a criticism of the thought that went into the policy choice [the ‘do nothing’ / Option 2 / Option 3 pivot] but in turning Option 2 into legal reality. The basis of liability between road users was settled over 100 years ago. The introduction of compulsory insurance nearly 90 years ago. The settled balance between liability and liability insurance is substantially disrupted here. That may be productive, but only if outcomes are deliberate and considered. By contrast, the issues I have identified with the Bill under consideration do not appear to be deliberate policy choices.

October 2017

[1] * Professor of Insurance & Commercial Law, Insurance Law Research Group, University of Southampton.

[1] As described in ‘Pathway to Driverless Cars: Insurance for Automated Vehicles’ IA No: DfT00366; RPC Reference No: RPC16-3522(1).

[2] Insurance Scholars Workshop, Rutgers Centre for Risk & Responsibility, Rutgers University, New Jersey (20th Oct 2017).

[3] This is an entirely hypothetical situation. I defer to sensor systems engineers for an accurate hypothetical.

[4] In Holden, a car which had failed its MOT was under repair by its insured owner in his employer’s garage. This was with the permission of the property owner. However, the welding caused the seat foam to catch fire and this spread to the garage and nearby buildings. The property insurers covering the property damage paid out, but sought to recover their losses (by subrogation) from the motor insurer. The motor insurer won at first instance on the basis that this was not ‘use’ of the vehicle and outside compulsory third-party liability cover under the 1988 Act. The Court of Appeal reversed this decision in August 2017.


Prepared 1st November 2017