Product liability and the issue of self-driving vehicles : is the EU system of product liability in need of a face lift ? (comparison with the U.S. system)

(1)

Hermance Schaerlig Thesis supervisor: Dr. J. Kortmann

Product liability and the issue of self-driving vehicles :

is the EU system of product liability

in need of a face lift ?

(comparison with the U.S. system)

(2)

As of 2012, road traffic accidents were the ninth leading cause of death in the world, and the World Health Organisation predicts that if no action is taken it could become the seventh by 2030. Luckily, the self-driving vehicle technology has the potential to change these alarming statistics. As it will eventually eliminate the driver element altogether and, because the overwhelming majority of the accidents stem from human error, the technology has the ability to reduce automobile accidents occurrences tremendously. However, legal uncertainty, principally about liability concerns, is susceptible to stifle innovation and to delay the development of this technology. And considering all the potential benefits it brings about, such as safety, environment, mobility/independence, and time allocation efficiency, it is crucial to provide a adapted legal framework and incentives to encourage the manufacturers to keep on developing the self-driving technology.

The lack of doctrinal thinking on the matter in the EU context drove me to examine the whether the EU system of product liability is in need of a reform as regards to the liability issues raised by the self-driving technology, and if so, which reform would better address the new technology?

I therefore conducted an advisory research. First, I defined the self-driving technology, describe its advantages and the potential liability issues it raises. Next, because the upcoming liability issues that the technology brings about have manly been examined by U.S. authors, in the context of the U.S. product liability system, I leaned on a comparison with the U.S. system of product liability. I established that, regarding concepts where autonomous vehicles are most likely to create complications, the two systems are quite similar substantially, but differ more on the procedure side. Regardless of such differences, after examining the systems shortcomings, I declared that they exhibit most of the same shortcomings. Therefore, I argued that the U.S. solutions are applicable to the EU system, and distinguish four different solutions: the “Adam Smith” solution, the adaptation of the current system of product liability, the new law option, and the compensation insurance fund. I showed that the first one does not fit in the current EU trend towards the autonomous technology. The second one involved to many hurdles for the self-driving technology’s development, and none of the related technologies provide any meaningful analogy. The third solution appeared to be more adapted, but there would be considerable procedural obstacles to a clear and efficient new law, and it would still risk hampering the technology’s development.

Therefore, I recommended the adoption of the fourth option: the non-fault insurance compensation fund, because it is the most optimal one socially, as all parties benefiting from the technology would split the cost of liability, because standardized awards involve many economic, practical and legal advantages, and because such scheme would encourage autonomous vehicle development without decreasing product safety, as it would still provide the right incentives toward the manufacturers. Moreover, even if there are potential obstacles to the implementation of that solution, they are likely to be overcome.

(4)

Introduction

As of 2012, road traffic accidents were the ninth leading cause of death in the world1, and the World Health Organisation predicts that if no action is taken it could become the seventh by 2030.2 Accordingly, safety is amongst the biggest concerns of the automobile industry, with approximately 102 billion dollars spent on research and development.3 Unfortunately, the precursory technological advances like airbags and seatbelts have failed to make most potentially fatal car accidents non fatal.4_{Luckily, a new cutting edge technology has the} potential to change these alarming statistics: the self-driving vehicle technology. Indeed, the overwhelming majority of the accidents stem from human error5, such as drunk-driving, distracted-driving or speeding. The self-driving technology will eventually eliminate the driver element altogether and has therefore the ability to reduce automobile accidents occurrences. Sebastian Thrun, one of the leading developers of Google’s self driving car, predicts that replacing all current cars with autonomous vehicles would reduce traffic accidents by as much as 90%.6

Its life saving potential combined with its expected profitability have made the self-driving technology a priority for car manufacturers and software companies: presently, at least nineteen car manufacturers are working on this technology as well as several software companies like Google, Microsoft or QNX7; in 2011, Toyota spent nearly 10 billion dollars on research into the self-driving technology8.

The Centre for Automotive Research predicts that the first commercially available, fully autonomous vehicles could be commercialized as early as 2019.9_{Indeed, Google cars have} successfully navigated jaywalking pedestrians, cars lurching out of hidden driveways, double-parked delivery vehicles, bicyclists who were not following street laws, and can even

1_{See Appendix 1.}

2_{WHO website, <www.who.int/mediacentre/factsheets/fs358/en/> accessed 21 April 2016.}

3_{Jeffrey R. Zohn, “When Robots Attack: How Should the Law Handle Self-Driving cars that Cause Damages”}

[2015] U. Ill. JL Tech. & Pol'y 2015 465.

4_ibid.

5_{Carrie Schroll, “Splitting the Bill: Creating a National Car Insurance Fund to Pay for Accidents in Autonomous}

Vehicles” [2014] Nw. UL Rev. 109 807.

6_{ibid 808.}

7_{Orly Ravid, “Don't Sue Me, I Was Just Lawfully Texting & Drunk When My Autonomous Car Crashing into}

You” [2014] Sw. L. Rev. 44 185-186.

8_{Ujjayini Bose, “Black Box Solution to Autonomous Liability” [2014] The. Wash. UL Rev. 92 1336.}

9_{Jack Boeglin, “The costs of self-driving cars: reconciling freedom and privacy with tort liability in autonomous}

(5)

recognize and respond to the hand signals of a police officer.10 California, Florida, Michigan and Nevada have already passed laws that allow the testing of autonomous vehicles on public roads.11 However, the quick development of this technology is already confronted with the slow pace of the laws governing responsibility for harm caused by the use or misuse of the emerging technology.12 Indeed, some commentators raise concerns that uncertainty over legal issues may be stifling innovation and delaying the development and deployment of this socially useful intelligent transportation system technology.13_{Thus, considering the} paramount societal advantages that this new technology brings about, the clarifications of the technology’s legal implication is of great priority.14 To be sure, autonomous vehicles will raise a myriad of new issues that may not be reconcilable with current regulations, specifically when determining the party responsible for causing a traffic accident, when an autonomous vehicle is involved.15 While some say that the current civil liability and automobile insurance systems are not designed to deal with the changes that autonomous vehicles will present,16_{others maintain that product liability law may be sufficient to govern} the introduction of autonomous vehicles technology.17 However, even the ones that believe the legal system is not nearly as hostile to autonomous cars as some may fear,18 agree that this technology will require revised legal requirements and important policy determinations.19 Indeed, it is crucial to create certainty for the market place.20 There are two main objectives to reconcile in this context: protecting consumers and fostering innovation.21 In other words,

10_{Zohn 481.}

11_{Boeglin 172.}

12_{Andrew R. Swanson, “Somebody Grab the Wheel: State Autonomous Vehicle Legislation and the Road to a}

National Regime” [2013] Marq. L. Rev. 97 2; Jaimee Lederman, Mark Garret and Brian D. Taylor, « Fault-y reasoning navigating the liability terrain in intelligent transportation systems » [2016] Public Works

Management & Policy 21 no 1 6.

13_{Lederman, Garret and Taylor 6; Kyle Colonna, “Autonomous cars and tort liability” [2012] Case Western}

Reserve Journal of Law, Technology & the Internet 4 no 4 84

14_{Kevin Funkhouser, “Paving the Road Ahead: Autonomous Vehicles, Products Liability, and the Need for a}

New Approach” [2013] Utah L. Rev. 437 459; Anders Eugensson, Mattias Brännström, Doug Frasher, et al. “Environmental, safety, legal and societal implications of autonomous driving systems” [2013] In: Proceedings

of the 23rd International Technical Conference on the Enhanced Safety of Vehicles (ESV), Seoul, South Korea. tractor using type-2 fuzzy logic controllers. Mechatronics, IEEE/ASME Transactions on. 287 2.

15_{Bose 1330; Zohn 463.} 16_{Schroll 806.}

17_{Swanson 4.}

18_{Dorothy J. Glancy, « Autonomous and Automated and Connected Cars-Oh My: First Generation Autonomous}

Cars in the Legal Ecosystem » [2015] Minn. JL Sci. & Tech. 16 2.

19_{Glancy 4.}

20_{Jeffrey K. Gurney, “Sue my car not me: Products liability and accidents involving autonomous vehicles”}

[2013] U. Ill. JL Tech. & Pol'y 272.

21_{Andrea Bertolini and Erica Palmerini, « Regulating Robotics: A Challenge for Europe » [2014] Upcoming}

(6)

making sure that manufacturers put reasonably safe products on the market22 as well as ensuring that liability is not such as to prevent the technology from being commercialized at all. In fact, all interested parties have a crucial interest in a transparent and carefully tailored regulatory environment of liability.23

Even though the European Union (EU) recently declared that it is working towards a coherent European framework for the deployment of automated driving,24 and mentioned liability as an issue to be addressed in this context, I found a relative lack of doctrinal thinking on the matter. This lack of insight drove me to examine the following question in this paper:

Is the EU system of product liability in need of a reform as regards to the liability issues raised by the self-driving technology? And if so, which reform would better address the new technology?

To reach an answer, I will conduct an advisory research and lean on a comparison with the U.S. system of product liability, as the upcoming liability issues that the technology will bring about have mainly been examined by U.S. authors, in the context of the U.S. product liability system. First, I will describe and define the self-driving technology, examine its advantages and the potential liability issues it raises. Next, I will compare the U.S. and the EU product liability systems, focusing on the concepts where autonomous vehicles are most likely to create complications. Then, I will compare the two systems shortcomings regarding the new technology, notice that they mainly face the same ones, and argue that the EU system of product liability is indeed ill-suited to address the self-driving technology and that, as a consequence, the technology’s development is in danger of being compromised. In order to remedy to this problem, I will claim that inspiration should be drawn from the ideas put forward by different U.S. authors. Subsequently, I will detail these solutions and consider them regarding the EU system. Then, I will argue that, amongst them, the most appropriate solution for the EU system is the establishment of a non-fault insurance compensation fund. And finally, I will conclude by summarizing my research and findings.

Before I begin I would like to mention that self-driving technology raises numerous other issues apart from the liability questions, such as ones regarding privacy25_{, freedom}26_or

22_{Gurney 271.}

23_{Bertolini and Palmerini 99.}

24_{Declaration of Amsterdam on cooperation in the field of connected and automated driving of the 14th of April}

2016, section I a)

25_{Sarah Aue Palodichuk, « Driving into the Digital Age: How SDVs Will Change the Law and Its}

(7)

international law27. However, I chose to focus on the liability matter, as it seemed to be the most relevant28 to address at that time.

26_{Boeglin 173.}

27_{Eugensson, Brännström, Frasher, et al. 6-8; Swanson 14.} 28_{Bertolini and Palmerini 112; Bose 1334}

(8)

I. Autonomous vehicles technology

i) Definition

For the purpose of this study, I will only consider the technology in its final phase of development, the “full automation stage”, which does not need a driver, but instead has all passengers performing other activities and not supervising the driving.29

As the Declaration of Amsterdam states that definition of automated driving should be based on the Society of Automotive Engineering (SAE) levels,30 I will only consider such standards. The SAE, makes a distinction between five levels of automation: 0, 1, 2, 3, 4, and 5, level 5 describing the full-time performance by an automated driving system of all aspects of the dynamic driving task under all roadway and environmental conditions.31

This paper will therefore address the liability issues raised by SAE’s level 5 of automation.

ii) History and characteristics

The development of this technology began in 2004 when the first autonomous vehicle competition, the Defence Advanced Research Projects Agency (DARPA) Grand Challenge, occurred.32 On this occasion, none of the vehicles were able to complete more than five percent of the race in autonomous mode.33 But at the next DARPA Grand Challenge, in 2005, five vehicles completed a 132 miles desert course while operating autonomously.34 Later, the DARPA held an Urban Challenge, which established that autonomous vehicles could safely operate in urban environments too.35

As one might guess, the self-driving technology is complex. Indeed, for autonomous vehicles to function, different types of advanced technologies need to work with each in performing all the necessary tasks.36 For example, the Google car consists of a structure on top of the car that navigates, detects traffic, and measures and analyzes the surroundings through the use of

29_{Eugensson, Brännström, Frasher, et al. 2.} 30_{Declaration of Amsterdam section II g)} 31_{Glancy 6-7.} 32_{Swanson 3.} 33_ibid. 34_ibid. 35_{Swanson 3-4.} 36_{Funkhouser 443.}

(9)

radar sensors, laser rangefinders, a video camera, a global positioning system (GPS) and maps.37 It is almost like the car has eyes.38

Such fully automated vehicles have the potential to provide significant social benefits in a wide range of areas.39 In addition to increasing road safety, they are also susceptible to positively impact the environment by increasing fuel efficiency and therefore reducing emissions.40 Furthermore, they have the unprecedented ability of enabling people who are handicapped, aged, impaired or otherwise unable to drive themselves to get around safely and independently.41_{Finally, the self-driving technology implies an increase in time allocation} efficiency, as people will be able to work or engage in another activity instead of driving.42 Despite all its advantages, the new technology brings about one fundamental issue, which the current product liability system is unable to address,43_{and that will inevitably require laws’} amendments to make its commercialization possible44: the question of liability assignment when an autonomous vehicle causes a road accident.

iii) Potential liability issues

Indeed, despite the fact that such vehicles will considerably reduce the occurrence of car crashes, as computers are not perfect, self-driving cars will sometimes malfunction and cause accidents.45 Because the commercialization of such vehicles will shift the responsibility of avoiding accidents from the driver to the vehicle itself,46 a number of liability issues will be raised. First, the development of such vehicles will involve a multitude of potentially liable parties, which will considerably complicate the assignment of liability between them. Secondly, despite the multitude of potentially liable parties, one can predict that the car or system manufacturer will remain the defendant of choice for victims of autonomous vehicles’

37_{Gurney 253.}

38_{see Chris Urmson’s TED talk at <www.ted.com/talks/chris_urmson_how_a_driverless_car_sees_the_road>.} 39_{Sophia Duffy and Jamie Patrick Hopkins, “Sit, stay, drive: The future of autonomous car liability” [2014] 16}

SMU Sci. & Tech. Law Rev. 122.

40_{Eugensson, Brännström, Frasher, et al. 9-10; Gurney 251.}

41_{Cohen Roy Alan « Self-Driving Technology and Autonomous Vehicles: A Whole New World for Potential}

Product Liability Discussion » [2015] Defense Counsel Journal 82 no 3 2.

42_{Zohn 470-471.}

43_{Duffy and Hopkins108.} 44_{Palodichuk 831.} 45_{Gurney 252.}

46_{Gary E. Marchant and Rachel A. Lindor, “Coming Collision between Autonomous Vehicles and the Liability}

(10)

malfunction.47 Consequently, it will be of primordial importance to clearly delimit the extent to which manufacturers can be held liable in order not to impede the technology’s development. Thirdly, because of their complexity, autonomous vehicles are sure to bring about more complications to liability litigations.48 Additionally, autonomous vehicles will also unravel procedural issues.49

All these liability concerns were in their majority brought forward by U.S. authors in the context of U.S. products liability laws and the U.S. market. To see if these issues are equally valid regarding the EU product liability regime, I will now compare the two systems and later examine the relevant shortcomings of both approaches in the context of the self-driving technology in a comparative perspective.

47_{Marchant and Lindor 3; Funkhouser 452; David C. Vladeck, “Machines without Principals: Liability Rules}

and Artificial Intelligence” [2014] Wash. L. Rev. 89 147-148.

48_{Glancy 17.} 49_{Gurney 273.}

(11)

II. EU and U.S. product liability systems

Both systems are based on a strict liability regime, which imposes liability on a manufacturer or seller, whose product causes injury, even though he has exercised all due care.50 Under both systems, liability is triggered when three cumulative conditions are met: a defect, a damage and a causal link between the two.51

The Product Liability Directive52 (the Directive) is binding on all EU Member States (MS). However, it does not mean that its principles are applied uniformly across Europe. Indeed, even though the Directive is of maximum harmonization, there remains interpretation leeway regarding many concepts, as there are not clearly defined, neither by the European Court of Justice (ECJ), nor in the Directive itself. In the ECJ’s defence, several Supreme Courts have not referred points to the ECJ, but have made unilateral decisions instead,53 bypassing its authority and therefore preventing it from harmonizing the interpretation of the Directive’s concepts. Moreover, many product liability cases are settled out of court by insurance companies,54 thus avoiding the judicial system. Additionally, the Directive is not applied very often. Indeed, because great uncertainty surrounds its concepts and application, and because it allows them to do so55, plaintiffs often prefer claiming damages on the basis of another system.

By contrast, the Restatement Third56 is a source of secondary authority; it is not binding on U.S. courts, but its influence in both resolving ongoing debates and predicting future legal

50_{Sara D. Schotland, “Overview of U.S. product liability regime” [2003] Arizona Journal of International &}

Comparative law 20 1.

51_{Taschner Hans Claudius, “Product liability: basic problems in comparative law perspective” in Fairgrieve}

Duncan, Product Liability in Comparative Perspective (Cambridge University Press 2005) 161.

52_{Council Directive 85/374/EEC of 25 July 1985 on the approximation of the laws, regulations and}

administrative provisions of the member States concerning liability for defective products (OJ 1985 L 210, p. 29) (see Appendix 2).

53_{Hodges Christopher, “Approaches to product liability in the EU and Member States” in Fairgrieve Duncan,}

Product Liability in Comparative Perspective (Cambridge University Press 2005) 195.

54_{Norbet Reich, Hans-W Micklitz, Peter Rott and Klaus Tonner, European Consumer Law (Intersentia 2014)}

245; Ivo Giesen, “The development of product liability in the Netherlands” in Whittaker Simon The development

of product liability (Cambridge University Press 2010) 168.

55_{Council Directive 85/374/EEC art. 13.}

(12)

trends within the U.S. is unquestionable.57 Indeed, although no state is required to adopt the Restatement, the majority of them have adopted many of its provisions.58

Thus, even if different on paper, the two instruments are quite similar practically. Therefore, I believe they are similar enough to be compared simultaneously for each concept that is particularly relevant in the context of the self-driving technology: producer, damage & compensation, defect, causation and defences.

i) Producer

The Directive provides a broad definition of the term producer,59 which is quite revolutionary as it brings about a considerable increase in the number of potentially liable parties.60_It includes the manufacturer of the product itself, the producer of any raw material or manufacturer of a component part, and any person who presents himself as the producer by putting his name, trademark or other distinguishing feature on the product.61 The notion also stretches to any person who imports into the Community a product for sale, hire, lease or any form of distribution in the course of his business.62 Moreover, the Directive provides that the supplier shall be liable as well unless he informs the victim within a reasonable time of the identity of the producer or his own supplier.63 Thus, the plaintiff no longer has to identify and sue the particular party who should have discovered the danger and addressed it accordingly, but can choose between all.64

The Restatement Third provides for a similar definition, the only difference is that distributors are usually shielded from strict liability under state law.65 Thus, contrary to the EU consumer, the American one cannot sue the seller if the manufacturer is not identifiable.66

57_{Rebekah Rollo, “Product liability: why the European Union doesn’t need the Restatement (third)” [2004]}

Brooklyn Law Review 69 2.

58_{Lauren Sterrett, « Product Liability: Advancements in European Union Product Liability Law and a}

Comparison between the EU and US Regime » [2014] Mich. St. Int'l L. Rev. 23 3.

59_{Patrick Kelly and Rebecca Attree, European product liability (Butterworths 1992) 5; Sterrett 5.} 60_{Kelly and Attree 157 and 284.}

61_{Council Directive 85/374/EEC art. 3(1).} 62_{ibid. art. 3(2).}

63_{ibid. art. 3(3).}

64_{Stapleton Jane, Product Liability (Cambridge University Press 1994) 243.} 65_{Sterrett 5.}

(13)

ii) Damage & compensation

The Directive does not provide any substantive rules on compensation but only guidelines, living the subject to the discretion of each MS.67 Thus, certain types of damage will or will not be compensated depending in which country the lawsuit takes place.

Regarding the American system, U.S. courts normally compensate pain and suffering on top of the usual damage.68_{But, similarly to the EU system, there is no clear and consistent theory} on how to compute such compensation, which can lead to awards that are disproportional to the harm suffered.69

Despite this apparent similarity, there is one big difference between the European system and the American one, that is that the latter allows punitive damages whereas under the former compensation is strictly compensatory.70 Even though recent decisions seem to indicate that the role played by punitive damages will decrease in future product liability litigations,71 the availability of potentially huge punitive damages still provides a strong incentive to initiate litigation or seek colossal amounts in settlements72.

iii) Defect

The issue of defectiveness has exercised the minds of legal scholars more than any other aspect of product liability law,73 and is perhaps the point on which the U.S. system and the EU one differ the most.

In Europe, the concept of defectiveness is based on a consumer expectations test.74 Indeed, a product is considered defective when it does not meet the safety expectations that a person is entitled to expect, considering all circumstances including: the presentation of the product, the use it could reasonably be expected that the product would be put, and the time when the product was put into circulation75_{. This definition has been very much criticised for being} circular, as what a person is entitled to expect is the very question a definition of defect

67_{Kelly and Attree 10; Council Directive 85/374/EEC art. 9.} 68_{Stapleton 286.}

69_{Lucas Bergkamp and Rod Hunter, “Product liability litigation in the US and Europe: Diverging procedure and}

damage awards” [1996] Maastricht J. Eur. & Comp. L. 3 403.

70_{ibid. 402-403; Schotland 1.}

71_{Andrew P. Garza, “Look Ma, No Hands: Wrinkles and Wrecks in the Age of Autonomous Vehicles” [2011]}

New Eng. L. Rev. 46 12.

72_{Bergkamp and Hunter 403.}

73_{C.-J. Miller and Richard Goldberg, Product liability (Oxford University Press 2004) 345.} 74_{Sterrett 5; Miller and Goldberg 352.}

(14)

should answer.76 Thus, it fails to provide a readily ascertainable objective standard against which a manufacturer or a court can measure the safety of a product.77

Some aspects of the concept have been defined. Firstly, the safety standard attached to the notion of defect is an objective one: the expectations to consider are the ones of the public at large and not of the concerned consumer.78 Secondly, the notion of “put into circulation” has been defined in the case Declan O’Byrne v Sanofi Pasteur, as referring to the moment at which the product is taken out of the manufacturing process operated by the producer and enters a marketing process in its final form.79_{Thirdly, consumers are typically not entitled to} expect absolute safety.80 They are only entitled to expect that a product complies with the standard of safety common to the items of a same line of products marketed by a particular manufacturer.81 However, other features of the concept are still left unclear.

For example, in the recent Boston Scientific case, concerning pacemakers that had an abnormally high risk of failure, the ECJ shed some light on the notion of defectiveness, but also brought about new questions.82_{The ECJ established that a product in the form of a} medical device implanted in the human body is already defective if products in the same product group have a significantly increased risk of failure.83 It held that the lack of safety doesn’t stem from the danger but from the abnormal potential for damage.84 As much ground breaking this decision seems to be, such statement brings about other complications, particularly because the notion of “abnormal potential for damage” would require a standard of comparison defining what a normal damage potential is.85 Moreover, the majority of authors doubt that this reasoning would be extended beyond medical products whose failure implies grave health consequences.86

76_{Stapleton 234.}

77_{Miller and Goldberg 354.}

78_{Council Directive 85/374/EEC 6}th_Recital.

79_{Declan O'Byrne v Sanofi Pasteur MSD Ltd and Sanofi Pasteur SA, C-127/04, ECLI:EU:C:2006:93, para. 24}

and 32.

80_{Reich, Micklitz, Rott and Tonner 249.} 81_{Miller and Goldberg 395.}

82_{Lucas Bergkamp, “Is there a defect in the European court’s defect test? Musing about acceptable risk” [2015]}

European Journal of Risk Regulation 6 12.

83_{Boston Scientific Medizintechnik GmbH v AOK Sachsen-Anhalt – Die Gesundheitskasse and}

Betriebskrankenkasse RWE, C-503/13 and C-504/13, ECLI:EU:C:2015:148 para. 41.

84_{Boston Scientific, Opinion of Mr Advocate General Bot delivered on 21 October 2014, para 30.} 85_{Bergkamp 3.}

86_{Barend Van Leeuwen and Paul Verbruggen, “Resuscitating EU Product Liability Law? Contemplating the}

Effects of Boston Scientific Medizintechnik GmbH v AOK Sachsen-Anhalt - Die Gesundheitskasse, and

(15)

Despite the EU’s attempts to clarify the notion of defect, a general lack of guidance on the proper interpretation of the concept of defectiveness still persist, which led to different approaches being taken by the MS. For example, regarding its definition: Belgian law defines the defect more broadly than the directive by using the words “unsuitable for use”87; the Italian version considers a product defective when it does not provide the same degree of safety as that normally offered by any other product of the same series88; the Polish regulation abandoned the notion of defect for the notion of lack of safety, but it follows the directive’s wording in stating that a product is unsafe if does not provide the safety which can be expected taking into account normal use of the product.89 National courts have also developed different practices in order to facilitate the proof of defect.90 Some continental courts have been prepared to accept that the mere fact that a product causes harm in an unexpected way can be the basis of liability.91 They consider unexpected damage caused by a product used as intended as evidence of a defect.92 Some, like Germany, Spain or the Netherlands, have adopted a system of reversal of burden of proof: when the product was used in a normal way, the fact that a damage occurs leads to the factual presumption that the damage had been caused by a defect, in other words a prima facie case of defect is established.93

By contrast, the §2 of Restatement Third provides a more comprehensive definition of defect than the EU Directive does, and thus eliminates much of the confusion prevalent in the latter.94 The Restatement Third has abandoned the doctrinal labels of strict liability and

87_{Kelly and Attree 67.}

88_{ibid. 229.}

89_{Sengayen Magdalena, “Product liability law in Central Europe and the true impact of the Product Liability}

Directive” in Fairgrieve Duncan, Product Liability in Comparative Perspective (Cambridge University Press 2005) 279-280.

90_{Geraint Howells, “Product liability – a history of harmonisation” in Fairgrieve Duncan, Product Liability in}

Comparative Perspective (Cambridge University Press 2005) 215.

91_{Geraint Howells, “Defect in English law – lessons for the harmonisation of European product liability” in}

Fairgrieve Duncan, Product Liability in Comparative Perspective (Cambridge University Press 2005) 148.

92_{Howells, “Product liability – a history of harmonisation” 215.}

93_{Stefan Lenze, “German product liability law: between European Directives, American Restatements and}

common sense” in Fairgrieve Duncan, Product Liability in Comparative Perspective (Cambridge University Press 2005) 114; Howells, Comparative Product Liability 131; Cees C. van Dam, “Dutch case law on the EU product Liability Directive” in Fairgrieve Duncan, Product Liability in Comparative Perspective (Cambridge University Press 2005) 131-132; Miquel Martin-Casals, “Spanish product liability today – adapting to the new rules” in Fairgrieve Duncan, Product Liability in Comparative Perspective (Cambridge University Press 2005) 50.

(16)

negligence and established separate functional definitions of liability.95 It distinguishes between three types of defect: manufacturing defect, design defects and warning defects.96

i manufacturing defect

Manufacturing defect designates a product, which departs from its intended design even though all possible care was exercised in its preparation and marketing.97

US courts apply strict liability to manufacturing defects in its purest sense by imposing liability even where the manufacturer has exercised all possible care when producing the product.98_{Moreover, even when the evidence that the product departs from the design is not} directly apparent, the plaintiff can still make a prima facie product liability case under the malfunction doctrine.99_{Indeed, the res ipsa loquitur doctrine allows inferring the existence of} a manufacturing defect when the product malfunctioned, the malfunction occurred during regular and proper use of the product, and the product was not altered or misused in such a way that could cause the malfunction.100

ii design defect

Design defects characterize situations where an entire product line meets the manufacturer’s design specifications, but raises the question whether these specifications create unreasonable risks.101 Design defects are considered to be the most problematic category of defects, because the concept of design may involve a range of factors,102 and because they have the potential to be much more widespread and can thus be of catastrophic consequences.103 The plaintiff used to need to establish whether the danger posed by the design is greater than an ordinary consumer would expect.104 However, as the consumer expectation criterion is often difficult to apply to complex products105, such as automobiles, the risk-utility test is now the dominant test,106 and consumer expectations are no more than a mere factor in the risk-utility

95_{Miller and Goldberg 351.} 96_{Schotland 1.}

97_{Restatement (Third) of Torts : Product liability §2a).} 98_{Miller and Goldberg 399; Funkhouser 446.}

99_{Restatement (Third) of Torts : Product liability §3.} 100_ibid.

101_{ibid §2b).}

102_{Miller and Goldberg 400.} 103_{Funkhouser 448.}

104_{Restatement (Second) of Torts § 402A (Am. Law Inst. 1965).} 105_{Sterrett 7.}

(17)

analysis107. Therefore, according to the Restatement Third, the plaintiff must prove that a reasonable alternative design existed at the time of the marketing and must show evidence that the manufacturer’s failure to adopt this alternative design made the product not reasonably safe.108 The unreasonably dangerous nature of the product should be determined taking into consideration all the following circumstances: usefulness and desirability of the product, safety aspects of the product, availability of a substitute product, manufacturer’s ability to eliminate the unsafe character of the product, user’s ability to eliminate the unsafe character of the product, user’s anticipated awareness of the dangers inherent in the product, and manufacturer’s ability of spreading the loss by setting the price of the product or by carrying liability insurance.109

First of all, the Restatement Third specifies that the alternative design must be reasonable and feasible.110 Then, in order to establish the existence of such alternative design, the crucial question is: what should be balanced against what?111 According to Miller and Goldberg, the risk-utility test is not a balance of the aggregate costs and benefits of the design itself, but rather, a micro level analysis of the costs and benefits of a manufacturer’s failure to alter the chosen design by adopting the plaintiff’s proposed design feature that would have prevented the plaintiff’s harm.112 Moreover, only reasonably foreseeable costs and benefits should be taken into consideration.113 If a product’s utility outweighs its costs, it will be considered as non-defective even if the risk is unavoidable.114

iii warning defect

The absence or inadequacy of warnings or directions for use may also very well render defective an otherwise acceptably safe product.115 Actually, failure to warn cases are the most common form of litigation under the U.S. product liability regime.116

Warning defects are defined along the same line as design defects. Indeed a product is defective in warning if the foreseeable risk of harm posed by it could have been reduced or avoided by the provision of reasonable instructions or warnings by the manufacturer and such

107_{Miller and Goldberg 418; Rollo 10.} 108_{Rollo 9.}

109_{Zohn 477.}

110_{Miller and Goldberg 409.} 111_{Miller and Goldberg 405.} 112_ibid.

113_ibid. 114_{ibid. 420.}

115_{ibid. 429 and 462.} 116_{ibid 430.}

(18)

omission rendered the product not reasonably safe.117 The adequacy of the warning depends on the type of product concerned: the greater the benefits, the greater the patient’s willingness to risk suffering some costs.118

It is not clear whether manufacturers are subject to any obligations to warn of subsequently discovered dangers.119 Usually, knowledge acquired after a product has been supplied will not generate an obligation to warn, although there remains a continuing obligation to take account of new information and rising standards as products of the original design continue to be supplied.120

As important as warning are, a too great reliance on them may have adverse effects, such as the overlooking of the defectiveness or otherwise of the product’s design, or decrease in the actual efficacy of warnings in situations where there are particularly necessary.121 Therefore, determining the ideal level of warning for a specific product can be quite challenging.

The Directive deliberately does not differentiate between types of defect; rather the definition is designed to include all types.122 However, that does not mean that it rejects such distinctions123, and some MS do differentiate: Germany, following the U.S. model distinguishes between manufacturing, design and warning defects124, English courts seem to differentiate between standard and non-standard defects125, as they distinguish between two different types of products, those which are and perform as the producer intended, and those which are different from what the manufacturer planned, because they are deficient or inferior in terms of safety126_.

If both systems seem to similarly treat the manufacturing defects, when considering design or warning defects, the EU privileges the consumer expectations test, whereas the U.S. has replaced it by the risk-utility test.

117_{Restatement (Third) of Torts : Product liability §2c).} 118_{Miller and Goldberg 439.}

119_{Miller and Goldberg 477.} 120_ibid.

121_{ibid 438.} 122_{Taschner 161.} 123_{Bergkamp 5.}

124_{Lenze 107; Howells, Comparative Product Liability 129.}

125_{Mark Mildred, “The development risks defence” in Fairgrieve Duncan, Product Liability in Comparative}

Perspective (Cambridge University Press 2005) 182.

(19)

iv) Causation

Causation is known to be one of the biggest problems of the strict liability regime instituted by the Directive.127 The Green Paper of 1999 already noted that, as the onus of proof rests with the claimant,128 causation might be a severe burden on the latter, when such proof turns out to be technically complicated and or expensive on account of the expert opinions required.129_{Some MS have developed instruments to overcome this issue}130_{, such as the} prima facie rule in Germany, which reverse the burden of proof onto the defendant131_. However, some still believe that such relaxation of the burden of proof is encouraging spurious claims and argue that it is upsetting the existing balance established by the Directive.132 Beyond the reversal of the burden of proof, some MS, like France, Belgium or Austria, seem to have accepted as sufficient to prove that the product had failed and that it resulted in an injury.133 Moreover, Spanish courts deemed the causation established when it has been duly established that their existence is more likely than not.134

Moreover, there is no harmonized concept of causation, and that it is left to the discretion of the twenty-eight different national laws.135 In the recent Boston Scientific136 decision, the ECJ did not provide any further guidance on the causation concept as it found the causal connection implicitly.137

Under the U.S. regime, the burden of proof also rests with the claimant. The causation issue is especially problematic with regard to warning defects, because the claimant must prove two causal links: he must establish a proximate causal link between the injury and the product allegedly lacking a warning exists,138 and must show that a warning would have altered the behaviour of those involved in the accident.139 However, practically, according to some, the

127_{Howells, “Product liability – a history of harmonisation” 209.} 128_{Miller and Goldberg 746.}

129_{Commission Green Paper: Liability for defective products, Brussels, COM (1999) 396 final (28 July 1999),}

3.2, 20.

130_{Sterrett 7.}

131_{Miller and Goldberg 748.} 132_{ibid. 749.}

133_{ibid 349.}

134_{Kelly and Attree 357.}

135_{Howells, Comparative Product Liability 196.}

136_{Boston Scientific, C-503/13 and C-504/13, ECLI:EU:C:2015:148} 137_{Bergkamp 11.}

138_{Miller and Goldberg 474.} 139_ibid.

(20)

problem is mitigated by the fact that juries tend to be easily convinced that a slight rewording or a more prominent warning would have made all the difference.140

Additionally, similarly to the Directive, the Restatement Third does not really address the causation concept and merely refers back to the prevailing rules and principles governing causation in tort, which can be freely interpreted by the courts.141

v) Defences

A number of defences are available under the Directive regime. However, in the context of self-driving vehicles, one particular defence appears to be of special relevance: the development risk or state of the art defence.

This defence takes unforeseeable risks outside the ambit of liability.142_{Indeed, the defendant} will not be held liable if he can prove that the scientific and technical knowledge at the time the product was put into circulation was not such as to enable the existence of the defect to be discovered.143 Accordingly the burden of proof rests on the defendant.144

Over the years some of the defence characteristics have been clarified by the ECJ. First of all, the concept of “state of scientific and technical knowledge” is an objective notion145, which refers to the overall sum of what scientists and technicians in their respective areas have achieved as common performance in their ongoing research and skills; the producer’s conduct and personal knowledge are irrelevant.146 Secondly, in Commission v United Kingdom, the ECJ stated that the defendant must prove that it was objectively impossible to discover the harmful characteristics of the product, taking into account the most advanced state of scientific and technical knowledge at the time the product was put into circulation, but specified that such knowledge must be accessible.147 Again the notion of accessibility is an objective one.148_{Thirdly, the degree of safety and availability of knowledge have to be} determined retrospectively at the moment when the product was put into circulation.149

140_{Miller and Goldberg 469.}

141_{Restatement (Third) of Torts : Product liability §15.} 142_{Stapleton 136.}

143_{Council Directive 85/374/EEC art. 7(e).} 144_{Kelly and Attree 8.}

145_{Commission v United Kingdom, C-300/95, ECLI:EU:C:1995:255, para 27.} 146_{Taschner 164.}

147_{Commission v United Kingdom, C-300/95, ECLI:EU:C:1995:255,}_{para 26.} 148_{Taschner 164.}

(21)

However some aspects remain unclear and have lead to different approaches to the defences being adopted by MS. For example, Austrian, German and English courts seem to exclude the defence in the case of manufacturing or non-standard defects.150 Moreover, according to Mildred, there is also a competition of interpretation amongst MS, which opposes the scholars relating the defence to a notion of “absolute undiscoverability” and the ones who speak of “undiscoverability by reasonable means”.151 In fact, the defence is torn between two competing objectives: the objective to provide the defendants with a real tool against excessive liability cases and the Directive’s objective of imposing strict liability onto the manufacturers.152 Overall, similarly to the notion of defect, the development risk defence concepts would need a clear interpretation from the ECJ in order to really harmonize the notion.

The U.S. system also knows of the development risk defence already from the time of the Restatement Second, whose comments j and k provided that manufacturer’s duty to warn was limited to foreseeable risks.153_{Nonetheless, in Beshada v Johns-Manville Products}

Corporation, the New Jersey Supreme Court held that the state of the art was a negligence defence and that unknowability of the dangers of asbestos was irrelevant in a strict liability claim.154 However, this decision was immediately subject to strong criticisms such as the potential uninsurability of manufacturer’s liability, as it is impossible to assess unknowable risks, or the disincentive to discover safety improvement for products already marketed155. Consequently, two years later the Supreme Court went back to the traditional negligence approach.156 Indeed, in Feldman v Lederle Laboratories, it held that the manufacturer would not be liable for failure to warn of risks, which were unknown and unknowable at the time of manufacture.157

The Restatement Third confirms this majority position by limiting the scope of responsibility to foreseeable risks of harm, and by stating that warnings should be within the capabilities of a prudent manufacturer, which seeks to make its product safe within the bounds of practically

150_{Howells, Comparative Product Liability 189; BGH, 1995, exploding mineral water bottle; OLG Frankfurt,}

1995, food infected with hepatitis A by a cook; Bergkamp 8; Simon Whittaker, “The development of product liability in England” 82.

151_{Mildred 170.} 152_{Stapleton 241.}

153_{Miller and Goldberg 508.} 154_{447 A2d 546 (NJ, 1982).}

155_{Miller and Goldberg 508 and 510.} 156_{ibid 511.}

(22)

available scientific knowledge.158 Accordingly, the conformance with state of the art is now only one element to take into account, and not an absolute defence.159

Similarly, the assumption risk defence, which provides that if the consumer purchased the product knowing its potential hazards, he assumed some or all of the responsibility for any harm that may follow the use of such product, has been merged into the comparative negligence analysis under the Restatement Third and is no longer recognized as a separate defence in many states.160

vi) Procedure

Even though the EU and the U.S. systems differ on some substantive law aspects, the major differences between the two regimes are procedural.161_{These differences are by no means} unique to product liability, but they are often more flagrant in this context.162

For the purpose of this chapter when I refer to the EU it will be at the exception of England, which exhibits for the most part the same procedural practices as the U.S.

When examining procedure, the two systems differ firstly in the frequency and in the way they use experts. The U.S. system makes an extensive use of party appointed experts, who some believe, are more at risk of being influenced or biased.163 Conversely, in the EU, party experts are used much less frequently, and court-appointed experts, who theoretically should be more impartial, are the rule.164 In fact, EU courts appoint independent experts, often chosen from a list of experts with authority in the local scientific community and experience in reporting to courts on technical matters.165

Secondly, under the EU product liability regime, key issues of the case are decided by judges who are appointed for life, have a law degree and some experience on the bench.166 By contrast, under the U.S. system, juries consisting of laymen decide the critical issues.167

158_{Miller and Goldberg 513.} 159_{Sterrett 9.}

160_{Marchant and Lindor 6-7.} 161_{Bergkamp and Hunter 417.} 162_{ibid. 400.}

163_{Schotland 2.} 164_{ibid. 400.} 165_{ibid. 407-408.} 166_{ibid. 401.}

(23)

Thirdly, there are non-negligible differences in the level of proof required for a finding of fault or defect between the two regimes.168 Indeed, in the U.S., the jury can (and is allowed to) find the product defective, even though the experts cannot show any basis for a reliable conclusion that a defect exists.169 Such controversial outcome is very unlikely under the EU regime. In fact, to overcome the fact that the jury is very much likely to find in the favour of the plaintiff when the latter provided barely any evidence to support his claim, U.S. defendants effectively bear the burden of proof.170_{Conversely, in Europe, the burden of proof} lies and often stays with the plaintiff.171_{Indeed, EU courts are not likely to shift the burden of} proof unless credible evidence is presented.172 Additionally, an important difference is the absence of discovery procedure in the EU. The U.S. discovery process allows the plaintiff to obtain copies of any documents that are in the defendant’s possession and vice versa.173 By contrast, the European party must show an overriding interest in the production of a particular document, and even so the granting of such request is subject to the court’s discretion.174 Fourthly, the last difference relates to the general costs associated with litigation. In Europe, the winning defendant is allowed to recover some litigation costs, which is not the case in the U.S.175 This impossibility has resulted in increased litigation and settlements providing for significant payments to claimants even if their claims appear to be wholly unsubstantiated; problems which are exacerbated in complex product liability litigations.176 Overall, the cost of defending a product liability lawsuit in the U.S. exceeds by many times the cost of defending a similar lawsuit in Europe.177 Indeed, significant awards and large number of claims have turned the U.S. product liability system into a multi-billion dollar industry.178 However, Rajneri still argues that litigation is already quite expensive in Europe, among others because of the absence of contingency fee and class action systems.179

168_{Bergkamp and Hunter 402.} 169_ibid. 170_{ibid. 412.} 171_{ibid. 414.} 172_ibid. 173_{ibid. 400.} 174_{ibid. 407.} 175_{ibid. 400.} 176_{ibid. 404.} 177_{ibid. 400.} 178_{ibid. 399 and 404.}

179_{Eleonora Rajneri, “Interaction between the European Directive on Product Liability and the former liability}

regime in Italy” in Fairgrieve Duncan, Product Liability in Comparative Perspective (Cambridge University Press 2005) 77-78.

(24)

III. Comparative shortcomings in the context of the self-driving

technology

From the description of the two systems, one can infer that they are quite similar at the substantial level, except perhaps when talking about the concept of defect and causation. However, because of their contrasting procedures and institutional and cultural settings it is relatively difficult to make strong extrapolations.180 I will nonetheless try to show that, despite their differences, the two systems share most of the shortcomings brought forward by the self-driving technology.

i) Producer

In many product liability cases and especially when the product is a vehicle, regardless of all the multitude of potential liable parties, the manufacturer will generally be considered as the target of choice181 because he has the deepest pockets182 and is the best placed to exercise control over the product’s quality and safety183.

Because autonomous vehicles are typically the product of many component parts coming from different manufacturers, the broad definition of producer takes particular relevance in the context of those vehicles. Indeed a malfunction in an autonomous vehicle will usually be a programming error or a system failure that could implicate several different potentially liable parties (manufacturer, manufacturer of a component part, software engineer, road designer…) and it may be hard to separate for the purpose of assigning responsibility.184 This multiplicity problem is more relevant under the Directive, which provides for a broader definition of producer than U.S. law does.

ii) Damage & compensation

Despite the mentioned differences regarding compensation for pain and suffering, and the availability of punitive damages, a common critique of legal uncertainty can be made. Indeed according to Garza, the randomness and uncertainty of punitive damages make it virtually

180_{Stapleton 68.}

181_{Marchant 1328; Vladeck 147-148} 182_{Marchant and Lindor 4.}

183_{Miller and Goldberg 212.} 184_{Marchant and Lindor 3-4.}

(25)

impossible for insurers and manufacturers to calculate their exposure.185 The same could be said for the EU system regarding the randomness of compensation, as each MS is free to determine compensation questions, and thus amounts can greatly vary depending on which national court in awarding them. Even if this problem of uncertainty is not limited to product liability law alone, it could be particularly relevant regarding the self-driving technology, because accidents involving autonomous cars are likely to entail important amounts of damages, as car injuries are often severe.

iii) Defect

i general remarks

The notion of defect will certainly prove to be one of the most problematic aspects of product liability law in the context of self-driving technology.

As previously seen, if the Directive does not differentiate between types of defects, some MS do and in fact the Directive’s definition does not exclude such distinction but rather includes all types.186 So I will use such differentiation to structure my analysis. Some authors argue that the manufacturing defect angle is very much equipped to address autonomous vehicles defects.187 However, a few concede that the malfunction theory gets significantly more complex when applied to autonomous vehicles.188 Thus, other authors take the view that the manufacturing defect angle is not adapted to address autonomous vehicles, as when concerned with a software error nothing tangible is in fact manufactured.189 Additionally, Marchant and Lindor claim that because autonomous vehicles have very low production error rates, manufacturing defects are unlikely to be applied very often to them.190 Therefore, authors agree that the majority of self-driving related claims would likely be brought under the design or warning defect angle.191

Determining the proper test for design and warning defects has been the central and most contentious question in products liability law for a very long time, and the issue does not appear to have been resolved.192_{I showed that the EU and the U.S. systems represent two} 185_{Garza 12.} 186_{Taschner 161.} 187_{Funkhouser 453; Cohen 3.} 188_{Funkhouser 455.} 189_{Cohen 3; Gurney 259.}

190_{Marchant and Lindor 1323-1324; Zohn 474.} 191_{Funkhouser 455-456.}

(26)

competing viable liability theories: the consumer expectations test and the risk utility test.193 Each theory presents its own advantages and disadvantages,194 but both entail significant complications in the self-driving technology context195.

ii inadequacy of consumer expectations

The consumer expectations test has been widely criticized as inadequate in cases of obvious dangers and foreseeable misuse, in cases where the claimant is not the original purchaser of the product, and in complex design defect cases.196 Indeed, when concerned with an obvious defect, the average consumer would be completely aware of the risk of harm,197_{and thus} could be presumed to have assumed such risk. In the context of complex products, consumer expectations have been said to be particularly misleading, as consumers may not really know what to expect because they do not fully understand the technology and can therefore have unreasonable expectations.198 Indeed, in Jackson v General Motors Corp., the court held that for complex products, such as fuel injection engines or air bags, an average consumer has no realistic way to understand how it functions.199 Furthermore, Owen argues that, already in the case of regular automobiles, consumers have no meaningful expectations as to the extent to which a vehicle may, or may not be compromised in the event of a collision at high speed.200 Indeed, some U.S. courts already hesitate to apply the consumer expectations test to most automotive accidents.201 Thus, one can expect them to be even more reluctant when it comes to autonomous vehicles related incidents, the technology involved being much more complex. Beyond the complexity of the product, the fact that it is a totally new technology also raises issues with respect to consumer expectations, as consumers do not have anything to base their expectations on.202 Additionally, the more novel and advanced the technology, the fewer the individuals who possess the requisite knowledge to have a reasonable expectation about the safety of the products.203 Considering the level of complexity of the autonomous vehicle intelligence system, one could presume that the reasonable consumer could expect a self-

193_{Vladeck 133-134.} 194_{Funkhouser 451.} 195_{Funkhouser 456.}

196_{Miller and Goldberg 357; Stapleton 235.} 197_{Miller and Goldberg 358.}

198_{ibid. 403; Funkhouser 456; Garza 5; Zohn 476} 199_{Jackson, 60 S. W. 3d at 804-05; Zohn 476.}

200_{David G. Owen, Product Liability Law (Thomson West 2008) 491.} 201_{Gurney 261.}

202_{Funkhouser 451.} 203_{Funkhouser 456.}

(27)

driving car to drive itself without mishap; but is such expectation really realistic?204 Some argue that indeed it is not and that therefore the consumer expectation test should not be used because consumers cannot expect perfection.205

Moreover, Stapleton argues that the consumer expectations test is not an appropriate standard in any tort liability system, because of the central role played by bystanders.206 According to her, it is unworkable to require a tort standard to be linked to some prior transaction with another party.207_{However, as the consumer expectations standard is an objective one, it is the} expectations of the public at large that are considered,208_{and thus one could argue that the} bystanders’ view is also taken into account. Moreover, some sustain that a bystander standard is not adapted for determining whether the product was unreasonably dangerous, as it would eliminate any base from which to determine unreasonableness, and the seller would become an insurer.209

For all these reasons, in the U.S., the consumer expectations test has been supplanted by the risk-utility approach for design and warning defects cases in the overwhelming majority of states.210 Some European MS too, like Germany, have acknowledged the former standard’s limitations and have combined it with other standards such as statutes, regulations and risk-utility comparison.211

iii problems with cost-benefit analysis

The risk-utility test also has its downsides. The technology’s complexity and novelty makes it more difficult to determine if an alternative design was available and would have provided better safety.212 Because the self-driving technology is so new, presenting evidence of an alternative design would be extremely difficult; it would require the involvement of an expert engineer and other experts, like economists, that would have to battle with the manufacturer’s experts,213 which would make it very expensive for plaintiffs to pursue claims214. Moreover, it will likely be a long time before there is a sufficient number of manufacturers involved in the

204_{Cohen 3.} 205_{Cohen 3.}

206_{Stapleton 234-235.} 207_{Stapleton 235.}

208_{Howells, Comparative Product Liability 36.} 209_{Miller and Goldberg 359.}

210_{Garza 5; Zohn 476.} 211_{Lenze 109-110.} 212_{Funkhouser 451.} 213_{Vladeck 138.} 214_{Cohen 3.}

(28)

production of autonomous vehicles such as to have substitute products to consider when trying to establish the availability of an alternative design.215

By contrast other authors claim that, in the case of autonomous vehicles, the risk-utility test provides a too easy avenue for plaintiffs. Marchant and Lindor argue that the manufacturer will indeed almost always lose the cost-benefit analysis when it focuses on the micro scale between slightly different versions of the autonomous system, because the cost of not implementing the potential improvement will usually be severe – the loss of one or more lives, or other serious injury compared to the relatively small cost of the marginal improvement that might have prevented the accident.216

iv) Causation

Beyond the notion of defect, causation is also one of the most problematic issues of product liability law, because of the significant hurdle it entails for the plaintiff. Indeed, it has been argued that causation might be a severe burden on the claimant, when proving the causal link between defect and damage is technically complicated or expensive on account of the expert opinions required.217 As autonomous cars are technically very complex, and will require the testimony of highly specialized expert,218 which can often be cost prohibitive219, causation is sure to represent an especially serious hurdle for plaintiffs in cases involving self-driving vehicles.

v) Defences

According to Cohen, the state of the art defence will prove to be of tremendous relevance in autonomous vehicle cases because this technology is developing in real time.220

I noted that contrary to the EU system under which the state of the art argument is a defence in itself, the U.S. regime does not recognize the development risk defence as such, but includes the argument in the general liability assessment221_{. However, in practice, as the} Restatement Third limits the scope of responsibility to foreseeable risks of harm, there is not

215_{Funkhouser 457; Zohn 477; Dylan Levalley, “Autonomous vehicle liability—Application of common carrier}

liability” [2013] Seattle UL Rev. 36 4

216_{Marchant and Lindor 5-6.}

217_{Commission Green Paper: Liability for defective products, Brussels, COM (1999) 396 final (28 July 1999),}

3.2, 20.

218_{Gurney 263.} 219_{Gurney 265.} 220_{Cohen 334.} 221_{Sterrett 9.}

(29)

much difference between the two regimes. Considering the relative uncertainties surrounding the defence’s concepts under the Directive, and knowing that U.S. defendants do not have to invoke the defence as such, one could infer that the risk development defence will prove more useful for U.S. defendants.

vi) Procedure

Additionally, autonomous vehicles will also exacerbate procedural matters already known. Firstly, the U.S. court system is usually criticized for its party-experts involvement. As they are hired, they potentially have more incentives to submit misleading opinions and biased scientific evidence; and the applicable rules allow them to do so.222_{One can expect that this} issue will be even more relevant in the context of self-driving vehicles, as the complex technology will likely force parties to heavily rely on experts’ testimony.

Secondly, it has been said that, in liability cases involving scientific or medical testimony, juries are not competent and tend to be partially.223 Some believe that juries are not fundamentally worse at dealing with factually and scientifically complex litigation than are judges,224 but that they are substantially more unpredictable.225 Moreover, they also say that juries appear to be significantly biased in favour of plaintiffs that face big corporations.226 Additionally, with the advanced technologies incorporated in autonomous cars, the nature of the evidence and experts is likely to make litigation especially challenging and complex technologically.227 Therefore, according to Marchant and Lindor, this potential issue is likely to be exacerbated, as there is some evidence that lay persons composing a jury are suspicious of unfamiliar and exotic-edge technologies, regardless of their actual potential of causing harm.228 The EU will not be confronted with this problem, as no jury is usually involved in civil lawsuits.229_{But it is not to say that EU judges are completely free from bias either.} Indeed, there could also be a risk of partiality in the EU courts.

222_{Bergkamp and Hunter 407.} 223_{ibid. 409-410.}

224_{ibid. 412.} 225_{ibid. 417.} 226_{Stapleton 75.} 227_{Glancy 17.}

228_{Marchant and Lindor 6.}

Product liability and the issue of self-driving vehicles : is the EU system of product liability in need of a face lift ? (comparison with the U.S. system)