Autonomy and the Law of the Faster Horse: In Response to the Adoption of Autonomous Vehicles, What Aspects of the Australian and Irish Legal Liability Frameworks Require Reform?

Pagespp 1 - 37
Published date12 January 2022
Date12 January 2022
1
Autonomy and the Law of the Faster Horse:
In Response to the Adoption of Autonomous
Vehicles, what Aspects of the Australian and Irish
Legal Liability Frameworks Require Reform?
CONNOR HOGG*
I. Introduction
Human beings have had an interest in creating tools that operate of their own
volition for much of recorded history. Indeed, the English word we use for such
technologies, automatic, is a Greek term. Mythological stories recorded by Homer
and Plato refer to Automatons, animate statues cast of bronze capable of working
for and even outsmarting their human creators.1 is drive to automate tasks, once
solely the domain of humans, has continued through to the modern era.
Not long aer their popular adoption, the rst attempts to create automobiles
that operate autonomously began to emerge by the 1920s.2 For nearly a century
these projects languished, with the technology to create a truly autonomous
vehicle (‘AV’) sorely lagging human imagination. However, computer processing
power, articial intelligence, imaging sensors, and internet connectivity have all
advanced an unprecedented amount in the 21st Century. Today, AV development
has become the primary research goal of not just technology companies but also
the companies with the largest market share in the automobile market such as
Ford.3 Despite never selling a single vehicle to a member of the public, Waymo is
already among the most valued automobile companies on the planet.4 As of 2019,
* Senior Analyst, Australian Competition and Consumer Commission; Adjunct Lecturer,
Macquarie University Law School; LL.M. (Dubl.), LL.B. (Macq.), B.A. (Macq.). e author would
like to acknowledge Mrs Ruth Tarlo, whose generosity in honour of her late husband Hyman
Tarlo, established a scholarship to support an Australian student undertaking postgraduate legal
studies at Trinity College Dublin. is article was the product of my academic research whilst the
Hyman Tarlo Scholar at Trinity College Dublin in 2018/19. Any views or opinions expressed in
this article are the personal views and opinions of the author.
1 e eoi Project, ‘Automatons (Automotones)’ ps://www.theoi.com/er/Automotones.
html> accessed 12 January 2019.
2 Adrienne LaFrance, ‘Your Grandmother’s Driverless Car’ (e Atlantic, 28 June 2016)
theatlantic.com/technology/archive/2016/06/beep-beep/489029/> accessed 22 March 2022.
3 Chris Ziegler and Nilay Patel, ‘Meet the New Ford, a Silicon Valley Soware Company’ (e
Verg e , 7 April 2016)
interview-electric-self-driving-car-soware> accessed 14 January 2019.
4 Graham Rapier, ‘Waymo is Worth $100 Billion More than Previous Estimates, Morgan Stanley Says’
(Markets Insider, 7 August 2018) nessinsider.com/news/stocks/google-
2  
AVs built by several companies operate without a back-up human driver. Waymo
AVs alone drive 1.6 million kilometres each month, many of these with ordinary
members of the public utilising the vehicles as a taxi service. ose bullish on the
prospects of AV technology make bold claims that most vehicle manufacturers will
shortly begin selling AVs to the public, with the target date constantly shiing.5
Others remain sceptical that the change will be as soon, instead viewing the shi to
AVs as likely to occur alongside future bans on the purchase of petrol and diesel cars
across major economies in the period between 2030 and 2040.6
It is easy to get caught up in the excitement that AVs promise. World Health
Organisation gures estimate that each year 1.35 million people die in vehicles.7 In
the US alone the costs relating to trac fatalities are over USD $260 billion, with
non-fatal injuries incurring an additional USD$365 billion.8 ere is also optimism
that AVs will increase productivity and mobility whilst decreasing congestion and
related environmental impacts.9 A future where trac operates at peak eciency,
with fewer cars on the road and fewer incidents occurring is undoubtedly an
enticing one. Nevertheless, a number of accidents involving AVs have occurred,
including a widely publicised fatal collision with a pedestrian in March 2018.10
Such incidents have sparked a conversation amongst citizens, technologists and
stock-price-waymo-worth-100-billion-more-than-before-morgan-stanley-2018-8-1027439248>
accessed 7 October 2018; Sam Abuelsamid, ‘Waymo’s $30 Billion Valuation Shows the New
Reality of Automated Driving is Sinking in’ (Forbes, 6 March 2020)
sites/samabuelsamid/2020/03/06/waymos-30b-valuation-shows-the-new-reality-of-automated-
driving-is-sinking-in/?sh=2f087a1f6f35> accessed 10 February 2022.
5 Alex Davies, ‘e WIRED Guide to Self-Driving Cars’ (WIRED, 13 December 2019)
www.wired.com/story/guide-self-driving-cars/> accessed 15 December 2019. See also Andrew J
Hawkins, ‘Tesla Privately Admits Elon Musk Has Been Exaggerating About ‘Full Self-Driving’
(e Verge, 7 May 2021) ttps://www.theverge.com/2021/5/7/22424592/tesla-elon-musk-
autopilot-dmv-fsd-exaggeration> accessed 10 February 2022.
6 Allianz Partners, ‘e World in 2040: e Future of Healthcare, Mobility, Travel and the Home’
(Allianz Partners, 2019) rs.com/content/dam/onemarketing/
awp/azpartnerscom/reports/futorology/Allianz-Partners-Mobility-of-the-Future-Report.pdf>
accessed 10 February 2022.
7 World Health Organisation, ‘Global Health Observatory: Road Trac Mortality’
int/data/gho/data/themes/topics/topic-details/GHO/road-trac-mortality> accessed 9 February
2022.
8 United States House of Representatives Committee on Oversight and Government Reform
Subcommittee on Information Technology, and Subcommittee on Transportation and
Public Assets, ‘“Internet of Cars” Hearing Testimony of Dean C Gareld, President and CEO
Information Technology Industry Council’
content/uploads/2015/11/11-18-2015-Joint-Subcommittee-Hearing-on-Internet-of-Cars-
Gareld-ITI-Testimony.pdf> accessed 15 December 2019.
9 ibid.
10 Ryan Randazzo, ‘Who was Really at Fault in Fatal Uber Crash? Here’s the Whole Story’ (e
Arizona Republic, 17 March 2019)
03/17/one-year-after-self-driving-uber-rafaela-vasquez-behind-wheel-crash-death-elaine-
herzberg-tempe/1296676002/> accessed 18 March 2019.
Autonomy and the Law of the Faster Horse 3
regulators regarding the need to consider how liability will be apportioned in the
AV context in order to reduce legal uncertainty for all stakeholders.11
is paper addresses how the strict product liability frameworks of Australia and
Ireland will interact with the issues raised by the widespread adoption of AVs.
While a great deal has been written about the situation in the United States, these
two countries have been selected as both are marketed as regional hubs with light
touch regulation keen to attract investment from AV technology companies.12 Both
nations currently lack comprehensive legal analysis of the liability implications of
AVs, and share a similar legal history in the product liability paradigm, transitioning
from regulation by case law to regulation by statute at around the same time.
Technological innovations are inherently interdisciplinary, and the legal framework
that governs them should be examined in a way that considers the social context of
the law.13 is paper will adopt a comparative socio-legal research methodology,
as understanding the existing liability framework and potentially necessary
reforms requires consideration of the economic and political context.14 Socio-
legal scholarship is benecial to examine how the law actually operates in the
technology sector, given the other factors that will inuence the adoption of AVs
and possible law reform.15 Limiting the inquiry solely to doctrinal analysis cannot
take into account the technological, economic and political considerations that
will underpin any regulation of the AV sector.16
Part II of this paper examines the current state of the AV industry and the
development of the legal liability frameworks that exist in Australia and Ireland.
Part III considers how the existing strict product liability systems in each country
would interact with a claim relating to an AV. Part IV examines the reform
discourse, and assesses their ecacy for the Australian and Irish legal systems.
11 Madeline Roe, ‘Who’s Driving at Car?: An Analysis of Regulatory and Potential Liability
Frameworks for Driverless Cars’ (2019) 60 Boston College Law Review 317, 345.
12 See for example Industrial Development Agency Ireland, ‘Why Ireland for Connected and
Autonomous Vehicles’ m/newsroom/publications/why-ireland-
for-connected-and-autonomous-vehicles> accessed 9 February 2022; Stephanie Richards,
‘Autonomous vehicle hub announced in South Australia’
au/industries/innovation/autonomous-vehicle-hub-planned-for-northern-adelaide/> accessed 9
February 2022.
13 eodore Hagelin, Technology Innovation Law and Practice: Cases and Materials (LexisNexis
2012) ix.
14 Darren O’Donovan, ‘Socio-Legal Methodology : Conceptual Underpinnings, Justications
and Practical Pitfalls’ in Laura Cahillane and Jennifer Schweppe (eds), Legal Research Methods:
Principles and Practicalities (Clarus Press 2016) 116.
15 ibid 110.
16 See for example Hugh Collins, ‘Law as Politics: Progressive American Perspectives’ in James E
Penner, David Schi and Richard Nobles (eds) Introduction to Jurisprudence and Legal eory:
Commentary and Materials (OUP 2002) ch 7.
4  
II. Law of the Faster Horse?
II.I. Historical background
In the late 19th Century, cities such as New York and London were experiencing
what was then referred to as ‘Great Horse Manure Crises.17 Urban planners were
forced to deal with tens of thousands of horses covering the streets in tonnes of
manure each day. It was envisaged that by the mid-20th Century, streets would
be ooded with manure nearly twice the height of the average person. is did
not come to pass, as while many fretted about the risks of horse-based transport,
inventors across the globe had already begun building the rst horseless carriages,
or what we now refer to as the car.
is crisis never eventuated as by 1912 there were more automobiles on the streets
of New York than horses, with the last horse drawn street car ending its service in
1917. is story is oen used as a parable, warning that by focussing on creating
a ‘faster horse’, one can overlook the sudden revolution that a technology like the
car can bring about.18 Automobiles created a fundamental shi in every aspect of
modern life. It is envisaged that AVs will do the same. Indeed, Davies notes that:
is 21st-century gold rush is motivated by the intertwined forces of
opportunity and survival instinct. By one account, driverless tech will add $7
trillion to the global economy and save hundreds of thousands of lives in the
next few decades. Simultaneously, it could devastate the auto industry and
its associated gas stations, drive-thrus, taxi drivers, and truckers. Some people
will prosper. Most will benet. Many will be le behind.19
Given the enormous social and economic impact driverless cars are anticipated
to have, parallels have been drawn to another equine based metaphor, the Law
of the Horse. is refers to the debate in the 1990s as the then nascent Internet
began to impact the law relating to contracts, property and crime. Two competing
views emerged, with then US Federal Court Justice Frank Easterbrook suggesting
that the ‘best way to learn the law applicable to specialised endeavours is to study
general rules’.20 In Easterbrook’s view, information technology law should be merely
conceived of as torts in the IT context, contracts in the IT context, and so on.21
17 Elizabeth Kolbert, ‘Hosed’ (e New Yorker, 16 November 2009) .com/
magazine/2009/11/16/hosed> accessed 18 November 2018.
18 ibid.
19 Davies (n 5).
20 Frank H Easterbrook, ‘Cyberspace and the Law of the Horse’ [1996] University of Chicago Legal
Forum 207, 207.
21 ibid, in which Easterbrook memorably states:
the best way to learn the law applicable to specialized endeavors is to study general rules. Lots of
cases deal with sales of horses; others deal with people kicked by horses; still more deal with the
licensing and racing of horses, or with the care veterinarians give to horses, or with prizes at horse
Autonomy and the Law of the Faster Horse 5
is view was vociferously opposed by Lawrence Lessig, who argued that certain
technological innovations such as the Internet create such a new paradigm for
society and the economy that a sui generis body of law should develop around it.22
For many, AVs are the next great shi in transportation-the faster horse. ey are
also a technological revolution that has raised similar questions regarding whether
it requires a distinct regulatory framework-the Law of the Horse. is paper
fuses these two concepts together, asking whether the existing legal framework
is adequate to regulate AVs, or whether it requires a Law of the Faster Horse, one
equipped to encourage innovation and maintain acceptable levels of consumer
protection.
II.II. e state of autonomous vehicle development
For the rst several decades of the 20th Century, AV ideas centred on the use of
magnetic tracks or radio impulses on roads to guide vehicles steering and speed.
ese inventions were piloted but never advanced into anything more substantive
than experiments or exposition gimmicks.23 Over time, semi-autonomous
technologies began to enter automobiles. Cruise control became increasingly
popular as a fuel saving mechanism in the 1970s.24 Semi-autonomous technologies
have now developed to the point that they can now ensure vehicles maintain speed
in relation to trac, stay in their lane and park themselves. ese Advanced Driver
Assistance Systems (‘ADAS’) remain the dominant form of automation presently
available to ordinary consumers, with the driver still maintaining ultimate control
over the vehicle.
AVs in their present incarnation were borne out of a 2004 competition organised
by the United States Government’s Defense Advanced Research Projects Agency
(‘DA RPA ’). Oering a million-dollar prize to any team able to build an AV capable
of navigating an 142 mile course, the most successful team managed a measly seven
miles. However, in the competition the following year, ve teams completed the
entire course. e DARPA competition is seen as the catalyst for the explosion in
AV research and development. AV companies like Google’s Waymo staed their
teams with former competition entrants, and dominant AV technologies, such as
the light detection and radar tools used to create 3D maps of an AV’s surroundings,
were rst built for the competition.25
shows. Any eort to collect these strands into a course on “e Law of the Horse” is doomed to
be shallow and to miss unifying principles.
22 Lawrence Lessig, ‘ e Law of the Horse: What Cyberlaw Might Teach’ (1999) 113 Harvard Law
Review 501.
23 Roe (n 11) 323.
24 Pennsylvania State University, Ward’s Auto World Volume 10 (Ward Communications 1974) 95.
25 Davies (n 5).
6  
While a full technical explanation of AVs are beyond the scope of this paper, it
is useful to clarify their basic operational functions. AVs operate through a series
of sense-plan-act processes, much like most robotic or telematic machines.26 e
operation of an AV can be broken down into six steps:27
A. Sense
1. A range of sensors and positional guidance systems create a map of the
road and the vehicle’s surroundings.
2. Laser light or radio-based detection systems identify obstacles.
3. Machine learning algorithms recognise obstacles, and track their motion.
B. Plan
4. GPS guidance and path planning algorithms determine the AV’s journey.
5. Decision making systems use machine learning to predict the likelihood
of incidents.
C. Act
6. e vehicle completes its manoeuvre.
e taxonomy of AVs is set out by the automotive standards body SAE International
and is set out in Figure 1.
Today, companies including Waymo run passenger ser vices at SAE Level 5.28 Tesla
CEO Elon Musk has claimed that all Tesla vehicles manufactured aer October
2016 have the technological capacity to operate at SAE Level 5 once deployed to
the vehicles in the future,29 a claim repeatedly challenged as being overly optimistic
and likely to mislead customers.30
In addition to the current state of AV technology, it is also necessary to explore the
existing and mooted business models in the industry. e dominant view within
the industry is that AVs will make private car ownership both too expensive and
unnecessary. Given the price to design and manufacture AVs, it is anticipated that
AV manufacturers or intermediary businesses will own the vehicles, with people
26 R Brooks, ‘A Robust Layered Control System for a Mobile Robot’’ (1986) 2(1) MIT AI Memo
864.
27 As adapted from Ben Lorica, ‘e Technology Behind Self-Driving Vehicles’ (O’Reilly, 6 October
2016) .com/ideas/the-technology-behind-self-driving-vehicles> accessed 15
December 2018. See also Michelle Sellwood, ‘e Road to Autonomy’ (2017) 54 San Diego Law
Review 829, 836-37.
28 Talon Homer, ‘Waymo Releases Guidelines for Autonomous Vehicle Tech Crash Response’
(e Drive, 20 October 2018)
autonomous-vehicle-tech-crash-response> accessed 22 October 2019.
29 Jordan Golson and Dieter Bohn, ‘All New Tesla Cars Now Have Hardware for “Full Self-driving
Capabilities”’ (e Verge, 19 October 2016)
tesla-autopilot-update-model-3-elon-musk-update> accessed 15 February 2019.
30 See Hawkins (n 5); PlainSite, ‘California Public Records Request Response: California DMV
Tesla Robo-Taxi / FSD Notess/28jcs0/california-dmv-
tesla-robotaxi--fsd-notes/> accessed 10 February 2022.
Autonomy and the Law of the Faster Horse 7
SAE
level Name Denition
Steering and
acceleration
/deceleration
Monitoring
driving
environment
Fallback
scenario
Autonomous
capacity
Human driver monitors the driving environment
0No
automation
e full-time performance by the human driver of all aspects
of a dynamic driving task, even when enhanced by warning or
intervention systems
Human
driver
Human
driver
Human
driver None
1Driver
assistance
e execution by a driver assistance system of either steering or
acceleration/deceleration using information about the driving
environment and with the expectation that the human driver will
perform all remaining aspects of the driving tasks
Human
driver
Human
driver
Human
driver
Some driving
modes
2Partial
automation
e driving mode-specic execution by one or more driver
assistance systems of both steering and acceleration/deceleration
using information about the driving environment and with the
expectation that the human driver will perform all remaining
aspects of the driving tasks
AV System Human
driver
Human
driver
Some driving
modes
AV System monitors the driving environment
3Conditional
automation
e driving mode-specic performance by an automated driving
system of all aspects of the dynamic driving task with the
expectation that the human driver will respond to intervene
AV System AV System Human
driver
Some driving
modes
4High
automation
e driving mode-specic performance by an automated driving
system of all aspects of the dynamic driving task, even if a human
driver does not respond appropriately to a request to intervene
AV System AV System AV
System
Some driving
modes
5Full
automation
e full-time performance by an automated driving system of
all aspects of the dynamic driving task under all roadway and
environmental conditions that can be managed by a human driver
AV System AV System AV
System
All driving
modes
Figure 1: SAE International’s Levels of Driving Automation for On-Road Vehicles
8  
hiring them only for the time they are needed.31 is is the model Waymo currently
uses for the self-driving taxi service it oers in Arizona. Similarly, existing vehicle
manufacturers such as Audi and Volvo already oer subscription services for users
to pay a monthly fee to use any vehicle from a range of models, changing them as
oen as they like.32
is approach aims to not only dramatically decrease the cost of using road
transport for consumers, but also completely remove the need for personal car
ownership. Common estimates are that cars spend 95% of their time not in use,
meaning each individual vehicle has an enormous amount of slack capacity that
could otherwise be more benecially used.33 By changing the paradigm of personal
car ownership to a model where people only use AVs when they need to travel, this
will cause a signicant decline in the number of vehicles on the road.34 Given the
wealth of trip data available to any AV company in this situation, they will also be
able to coordinate trips to allow passengers travelling to similar locations to pool
together, further reducing the cost and need for vehicles.
In the event that AVs upend the existing transportation industry, the design of
the vehicles and the business model upon which they will operate require legal
systems across the globe to question the adequacy of existing frameworks. Some
jurisdictions-most notably the United States-have begun to put serious thought
into how AVs at SAE Level 3 and above should be regulated.
II.III. e development of existing legal liability frameworks
Developments in product liability can be traced to follow innovations in
technology and commerce. Until the late 19th Century, product liability in tort law
was conned to a very narrow set of situations where there was a sucient nexus
between the contracting parties and the harm that occurred. Privity of contract
or the existence of express warranties guaranteeing the quality of the product
remained the determinant factors.35 Winterbottom v Wright stands as the exemplar
of this period for vehicle related liability claims,36 with an injured driver unable to
claim aer being involved in a crash as the purchase agreement was between the
horse drawn carriage’s manufacturer and the driver’s employer.
31 Carrie Schroll, ‘Splitting the Bill: Creating a National Car Insurance Fund to Pay for Accidents in
Autonomous Vehicles’ (2015) 109 (3) Northwestern University Law Review 801.
32 Audi Select accessed 9 February 2019; Volvo, ‘Care By Volvo
pping-tools/purchase/care-by-volvo> accessed 9 February
2019.
33 e Economist, ‘e Perilous Politics of Parking’ (e Economist, 6 April 2017)
economist.com/leaders/2017/04/06/the-perilous-politics-of-parking> accessed 9 February 2019.
34 Carrie Schroll (n 31), 805.
35 See for example Chandelor v Lopus (1603) 79 ER 3; Crosse v Gardner (1688) Carthew’s King’s
Bench Reports 90; Pasley v Freeman (1789) 100 ER 450; Longmeid v Holliday (1851) 155 ER 752.
Autonomy and the Law of the Faster Horse 9
At the turn of the 20th Century, the ad hoc decision making of judges in the common
law had created a legal framework that was too cumbersome and impractical for
businesses trying to operate in an era of rapid technological advancement and mass
production. Owen observes that the rst half of the 20th Century can be seen as
a period of stretching legal principles to achieve desired results, with technology
rapidly outpacing law.37 Courts struggled to adapt old precedents to a completely
new technological and economic paradigm. e automobile sparked a signicant
rise in the number of liability claims and with it the slow growth of what we now
understand to be product liability law.38 Future US President Richard Nixon aptly
noted in an article whilst a law student that automobile liability law went from
being non-existent at the beginning of the century to being compendious enough
to ll an encyclopaedia by the 1930s.39
e late 20th Century saw the evolution of product liability law into what it is
today. Both Ireland and Australia passed comprehensive legislation to address the
issue-the former the result of the EU Product Liability Directive (‘PLD’) and the
latter inspired by it.40 ese laws introduced new principles to operate alongside
the common law, ensuring consumers in an increasingly globalised economy were
able to seek eective remedies in the event of harm caused by a defective product.
Part III of this paper assesses the ecacy of these existing legal frameworks and
their interaction with AVs.
III. New Technologies, Old Laws
III.I. A primer on the Australian and Irish product liability law
An exhaustive recitation of the law on product liability in both Ireland and
Australia is not possible given the strictures of a paper. Nonetheless, it is necessary
to provide at least a brief outline of the frameworks in each jurisdiction.
Ireland’s primary piece of product liability legislation is the Liability for Defective
Products Act 1991. e Act is Ireland’s statute implementing the PLD, which
creates a harmonised framework for product liability law across the EU that
37 David G Owen, Products Liability Law (3d edn, omson Reuters 2015) 249.
38 ibid. Se e also MacPherson v Buick Motor Company (1916) 217 NY 382 (New York Ct App);
Donoghue v Stevenson [1932] UKHL 100, 1932 SC (HL) 31; Kirby v Burke and Holloway [1944]
IR 207; Grant v Australian Knitting Mills (1933) 50 CLR 387. ese cases relate to the mass
market production of automobiles, carbonated beverages, and underwear respectively.
39 Richard M Nixon, ‘Comment: Changing Rules of Liability in Automobile Accident Litigation
(1936) 3 Law & Contemporary Problems 476, 476.
40 Defective Products Act 1991 (Ireland); Council Directive 85/374/EEC on the approximation of
the laws, regulations and administrative provisions of the Member States concerning liability for
defective products [1985] OJ L210/29 (PLD); Competition and Consumer Act 2010 (Australia)
which was previously titled the Trade Practices Act 1974.
10  
Member States cannot depart from.41 Consumers may also rely on rights under
the law of contract (as augmented by the Sale of Goods Act 1893 and the Sale of
Goods and Supply of Services Act 1980) or the tort of negligence.
Under the PLD, a product is understood to be defective if it does not meet the
‘legitimate expectations of the general public’ in regard to the safety of the product.42
Factors to be taken into account to determine these expectations include:
a) the presentation of the product;
b) the use to which it could reasonably be expected that the product would
be put;
c) the time when the product was put into circulation.43
Manufacturers (referred to as ‘producers’ in the PLD) are strictly liable for damage
caused by their products, be it physical injury or damage to property other than the
defective product itself.44 e ambit of who can be considered a manufacturer under
the PLD is broad and can include component manufacturers, persons branding
white label goods as their own, or even importers.45 Strict liability only requires
the injured person to ‘prove the damage, the defect and the causal relationship
between defect and damage’.46 e most relevant defence to strict liability for new
technology products is the development risks/state of the art defence, that protects
a manufacturer from liability if ‘the state of scientic and technical knowledge at
the time when [they] put the product into circulation was not such as to enable the
existence of the defect to be discovered’.47
Product liability law in Australia is governed primarily under the Competition
and Consumer Act 2010. Schedule 2 to this Act is the Australian Consumer Law
(‘ACL’). is codies all aspects of consumer protection legislation nationally.
Under Australia’s federal structure, all State and Territory governments have passed
legislation referring to the ACL as the relevant statute for consumer law issues in
their jurisdiction. Consumers can also rely on actions under the common law tort
of negligence or contract, although manufacturers are not strictly liable under these
non-statutory actions.48 e ACL contains a number of statutory warranties and
41 Case C-183/00 Sánchez v Medicina Asturiana SA [2002] ECR I-03901; Case C-52/00
Commission of the European Communities v French Republic [2002] ECR I-03827.
42 PLD art 6; implemented through s 5 of the Liability for Defective Products Act 1991 (Ireland)
s5; A v National Blood Authority [2001] 3 All ER 289 considering Consumer Protection Act 1987
(UK) s 45(1).
43 PLD art 6. is was implemented in Ireland through s 5 of the 1991 Act.
44 PLD art 1; L iability for Defective Products Act (Ireland) s 2.
45 PLD art 3; L iability for Defective Products Act (Ireland) s 1.
46 PLD art 4; L iability for Defective Products Act (Ireland) s 4.
47 PLD art 7(e); Liab ility for Defective Products Act (Ireland) s 6(e).
48 Adrian Coorey, Australian Consumer Law (LexisNexis Butterworths, 2015) 584–86; Colin
Loveday and Andrew Morrison, ‘Guide to Product Liability Laws and Regulations, Australia
(International Comparative Legal Guides, 2019)
liability-laws-and-regulations/australia> accessed 4 March 2019.
Autonomy and the Law of the Faster Horse 11
guarantees, including that goods are of acceptable quality, are t for any stated
purposes, and correspond with their description.49 Any end user of the product
(referred to as a ‘good’ in Australia) can seek relief if they suer loss or damage
due to their use of the product. Similar to the principles of the PLD, strict liability
applies to manufacturers and the manufacturer can encompass all actors along the
supply chain.50
In order for products to be of acceptable quality under the ACL, the product must
be free from defects.51 Under section 9(1), products ‘have a safety defect if their
safety is not such as persons generally are entitled to expect’. Much like European
jurisprudence related to the PLD, Australian courts consider the standards of the
general public rather than the specic knowledge and expectations of the plainti.52
Factors relevant to determining the expected level of safety include:
a) the manner in which, and the purposes for which, they have been marketed;
and
b) their packaging ; and
c) the use of any mark in relation to them; and
d) any instructions for, or warnings with respect to, doing, or refraining from
doing, anything with or in relation to them; and
e) what might reasonably be expected to be done with or in relation to them;
and
f ) the time when they were supplied by their manufacturer.53
As the doctrine of privity does not apply, any ultimate end user of (or any other
person suering injury or damage caused by) the defective product may make
a claim under the ACL. is includes for personal injury or death; claims by
their dependents; damage to other products as a result of the defective product;
or damage to land, buildings or xtures.54 e development risks/state of the art
defence in the PLD is directly mirrored in the ACL.55
III.II. Component liability
e modern automobile is not a purely analogue product. e components of
human operated vehicles manufactured in 2019 already include a plethora of
complex electronics and inbuilt computer systems. AVs are to be no dierent,
made up of component parts from numerous manufacturers. ese include a
49 ACL pt 3–2.
50 ibid ss 7, 138–42.
51 ibid s 54.
52 Loveday and Morrison (n 48). See also A v National Blood Authority [2001] 3 All ER 289.
53 ACL s 9(2).
54 ibid ss 138–41.
55 ibid s 142(c).
12  
mechanism for passengers to interact with the vehicle, sensor technologies for the
vehicle to capture data about its surroundings, automated vehicle controls, and an
articial intelligence system capable of operating the vehicle.56 A range of existing
component manufacturers and new market entrants are already competing to
manufacture the parts for the current prototype AVs being piloted (see g 2).
Cameras &
LIDAR
Mobileye
Velodyne
Val e o
Car
interiors
Continental
Bosch
Frim o
Car
exteriors
Roush
Delphi
Edison2
Chipsets
Nvidia
ualcomm
Samsung
Retrotting
systems
Cruise
Zoox
nuTonomy
Figure 2: A sample list of the original equipment manufacturers (OEMs) currently working
on providing AV components57
Existing product liability law in both Australia and Ireland is equipped to consider
defects in relation to the new hardware on and in AVs. Defects relating to physical
components or parts in automobiles are the same that vehicle manufacturers
currently experience. Both the PLD and the ACL have determined that
manufacturers are to be strictly liable for these types of defects as they are best
placed to engage in quality control testing and spreading the risk across all sales.
Legal uncertainty does not arise due to the fact that dierent components and parts
are used in AVs.58 Instead, the uncertainty exists in relation to driving behaviour
decisions made by autonomous systems. In his magnum opus on the act of driving,
Tom Vanderbilt aptly opined that:
e road, more than simply a system of regulations and designs, is a place
where many millions of us, with only loose parameters for how to behave,
are thrown together daily in a kind of massive petri dish in which all kinds of
uncharted, little-understood dynamics are at work.59
56 Dorothy J Glancy, ‘Autonomous and Automated and Connected Cars-Oh My! First Generation
Autonomous Cars in the Legal Ecosystem’ (2015) 16 Minnesota Journal of Law, Science &
Technology 619, 634 (2015). Mark Bergen, ‘Meet the Companies Building Self-Driving Cars
for Google and Tesla (and Maybe Apple)’ (RECODE, 27 October 2015) vox.
com/2015/10/27/11620026/meet-the-companies-building-self-driving-cars-for-google-and-
tesla> accessed 14 December 2018.
57 Adapted from Mark Bergen, ‘Meet the Companies Building Self-Driving Cars for Google and Tesla
(and Maybe Apple)’ (RECODE, 27 October 2015) .vox.com/2015/10/27/11620026/
meet-the-companies-building-self-driving-cars-for-google-and-tesla> accessed 14 December 2018.
58 Mark A Geistfeld, ‘A Roadmap for Autonomous Vehicles: State Tort Liability, Automobile
Insurance, and Federal Safety Regulation’ (2017) 105 California Law Review 1611, 1633.
59 Tom Vanderbilt, Trac: Why We Drive the Way We Do (And What It Says About Us) 6 (Vintage
2008).
Autonomy and the Law of the Faster Horse 13
It is in this proverbial petri dish where human error is currently the cause of almost
all road accidents. AVs upend this paradigm, as driver error is eliminated when
humans are no longer in control of their vehicles. However, the question must
then turn to whether errors in the design or programming of AVs that lead to harm
constitute defects for the purposes of product liability law.60
III.III. e inuence of European Union law on the Irish position
Compared to the other large jurisdictions, most notably the United States,
the European Union has been far more willing to legislate in relation to new
technologies.61 is willingness to play a more interventionist role is not without
its critics, with the majority opinion in many industries being that EU law creates
frameworks that impede innovation.62 is widely held belief exists in stark
contrast to the EU’s claimed aim to stimulate innovation and growth, particularly
in the elds of robotics and autonomy.63
At present, there is no European framework addressing specically motor vehicle
incident liability issues. e European Commission’s attempts in 2002 for a
harmonised framework failed, with the only relevant Directive being that related
to motor vehicle insurance.64 As such, despite the Commissions Europe 2020
strategy aiming to catalyse innovation in the AV sector through harmonised
laws, the existing EU framework is inadequate to achieve this aim.65 e PLD is
the relevant legislation that governs liability issues regarding the manufacturers,
retailers, and consumers of defective and unsafe products across the EU.66 Much
like other similar directives, the PLD aims to harmonise consumer protection and
create a level playing eld across the single market. e PLD primarily does this
by prohibiting the limitation or waiver of consumer rights to seek remedies for
damage caused by defective goods, and the no-fault liability scheme that applies to
manufacturers.67
60 Geistfeld (n 58) 1633.
61 Roeland de Bruin, ‘Autonomous Intelligent Cars on the European Intersection of Liability and
Privacy Regulatory Challenges and the Road Ahead’ (2016) 3 European Journal of Risk Regulation
485, 492.
62 Commission, ‘ird Report on the Application of Council Directive on the approximation of
laws, regulations and administrative provisions of the Member States concerning liability for
defective products (85/374/EEC of 25 July 1985, amended by Directive 1999/34/EC of the
European Parliament and of the Council of 10 May 1999)’ (2006) COM 496 nal, s 4.
63 Communication from the Commission, ‘Europe 2020, A Strategy for Smart, Sustainable and
Inclusive Growth’ COM(2010) 2020 nal.
64 de Bruin (n 61), 492; Directive of the EU Parliament and the Council 2009/103/EC Relating to
insurance against civil liability in respect of the use of motor vehicles, and the enforcement of the
obligation to insure against such liability.
65 de Bruin (n 61) 486.
66 Council Directive 85/374/EEC (n 41).
67 de Bruin (n 61) 490.
14  
e PLD imposes a no-fault liability structure on manufacturers that would be
applicable in the AV sector. is doctrine aims to enhance the interests of consumers
by creating an incentive for manufacturers to adequately consider product safety
issues.68 However, the Commission has acknowledged that the procedural aspects
of the PLD undermine the Directive’s objective. In a 2006 report, the Commission
noted that:
consumers are unfairly disadvantaged by the burden of having to prove
defect and/or causation in product liability claims. e concern mainly arises
from perceived diculties in proving claims due to a lack of legal or other
resources needed to investigate them properly, or to an inability to gain access
to essential information. Such problems are seen to be particularly acute in
relation to highly technical products, or where the alleged injuries are of a
complicated nature.69
AVs are a quintessential example of a product likely to become widely used, but
where the product safety issues and defects are technically complex. As this paper
has noted earlier, AV liability issues will force courts and regulators to consider
dicult questions. An incident involving an AV may require consideration of when
the car is beyond the driver’s control, when the hardware has caused the incident,
and when the soware’s learning algorithms caused the car to act in a way that may
make sense from previous experience, but was inappropriate in the exact situation.
It appears product liability claims relating to AVs will be among the most dicult
for ordinary consumers.70
Like many nascent technologies, regulation in the realm of AVs is a dicult
balancing act. Regulating too quickly can stymy an industry before it has a chance
to develop. A dearth of regulation on the other hand can also act as a disincentive,
as legal uncertainty may hold back experimentation. e same can be said for the
allocation of risk and liability under such regulations. Without liability risk on
manufacturers, consumers are more reluctant to adopt a new technology; while too
great a liability burden may ease consumer concerns, but disincentivise the release
of such technologies.71 While the existing PLD may strike an eective balance
between these competing policy objectives, its procedural aspects are a signicant
barrier to adequately protecting consumers who suer harm caused by AVs.
III.IV. e appropriateness of product liability legislation in the AV context
e very nature of AVs also raises a dierent set of product liability considerations
than traditional components of an automobile. A car’s brakes or its airbags are, what
68 ibid 488–89.
69 Commission (n 63) p 9.
70 de Bruin (n 61) 486.
71 Kirsty Horsey and Erika Rackley, Tort Law (OUP 2013) 363.
Autonomy and the Law of the Faster Horse 15
is referred to in the parlance of computer science, a deterministic system.72 An airbag
or set of car brakes do not make decisions of their own accord, and there is only a
singular predetermined outcome. In the case of an airbag, a percussive force (such as
a crash) triggers a sensor in the airbag, igniting a chemical reaction to ll with gas in
a matter of milliseconds.73 In a scenario where the expected outcome does not occur,
such as the recent mass recall of Takata airbags in millions of vehicles, the product
liability framework is applicable as this is almost certainly a defect.74
An AV on the other hand is programmed with a series of rules to countless
potential scenarios; the vehicle itself is the ultimate decision maker based on the
environment it nds itself in. A 2015 Google report on their self-driving car project
observed that ‘teaching a self-driving car to handle every possible situation it could
encounter on the road is not feasible, as there’s an innite number of possibilities.75
e report pointed to the example of one of their vehicles encountering a dog on
a leash dragging a person on a skateboard. While the vehicle did not contain any
algorithms for this specic situation, the technology has ‘fundamental capabilities
to respond correctly to unexpected situations as they happen.76
One of the key benets of autonomous technologies is the ability to adapt. Utilising
machine learning an AV-indeed even a eet of AVs-can begin to make better
decisions aer learning from their previous experiences with environmental factors
and other road users. is creates a situation where the vehicle’s programming
and decision-making processes are not set solely by the manufacturers during the
production process. is makes it more likely that an AV may cause an accident
based on a decision it learnt based on previous experience, rather than its original
programming.77
For Zipp, this means AVs can no longer be viewed as products in a traditional sense
of the term. His analysis is based on the view that although algorithms governing
the AV’s actions may cause an accident, this is not inherently evidence of a defect.
Product liability law was not draed to consider whether a decision taken by a
product that can make independent decisions can amount to a defect. Zipp notes
that to do so would require manufacturers ‘to conceive of every possible situation
the vehicles might encounter prior to their release’.78
72 John W Zipp, ‘e Road Will Never Be the Same: A Re-examination of Tort Liability for
Autonomous Vehicles’ (2016) 43 Transportation Law Journal 137, 163.
73 National Highway Trac Safety Administration, ‘Air Bags’ .nhtsa.gov/equipment/
air-bags> accessed 11 November 2018.
74 ibid.
75 Google, ‘Goog le Self-Driving Car Project Monthly Report July 2015’
content.com/media/www.google.com/en//selfdrivingcar/files/reports/report-0715.pdf>
accessed 5 November 2018.
76 ibid.
77 See Andreas Matthias, ‘e Responsibility Gap: Ascribing Responsibility for the Actions of
Learning Automata’ (2004) 6 Ethics and Information Technology 175, 177-82; Emad Dahiyat,
‘Intelligent Agents and Liability: Is It A Doctrinal Problem or Merely a Problem of Explanation?’
(2010) 18(1) Articial Intelligence Law 103, 106.
78 Zipp (n 72) 163.
16  
ough academic thought on the issue remains contested, jurisdictions such as
the EU appear to have preliminary views on whether soware and algorithms are
products within the ambit of the PLD or are services.79 e law remains uncertain,
with the Commission keen to clarify what is within the scope of the PLD without
overextending the reach of the Directive in a way that would unduly regulate
services. However, the Commission’s view on AVs would likely be that they view
it as a product. ey reason that the decision-making algorithms embedded into a
physical product ought to constitute a component of the product.80
If this analysis is accepted, the technical complexity creates a dicult evidentiary
and procedural burden for consumers, as noted in Part III.III above.81 Proving that
the decision-making algorithms are a defective product relies on establishing the
reasonable safety expectation of consumers.82
III.V. Defectiveness determinations
Determining the expectations of a consumer in the autonomous car context will
require courts to consider the feasibility, practicality, and costs considerations
associated with autonomous car design. e problematic manner in which complex
technical products are evaluated is likely to eventuate in two outcomes. Under US
law, a consumer complaint may fail as their expectations of the vehicle’s algorithm
capabilities may be technically unfeasible or impractical, or prohibitively costly.83
Conversely, US courts may be willing to nd AV producers liable for the majority of
incidents caused by the vehicle, without adequate consideration of the technically
complex questions of whether there is a safer alternative design that is suciently
feasible and practical.84 Both outcomes are certain to harm a nascent AV market.
Despite the unique nature of autonomous technologies, for the time being it
appears almost certain that AVs will fall under the ambit of existing product
liability law. Analysis must then turn to considering when an AV manufacturer is
liable under the law. e test under both the PLD and the ACL is controversial,
particularly in its applicability to technically complex products.85 e PLD and the
ACL do not automatically hold the manufacturer at fault when harms are suered
as a result of their products. e key causative factor to trigger the strict liability
provisions in both laws is the requirement that the product be ‘defective’.86 Both
79 European Commission, ‘Workshop on liability in the area of autonomous systems and advanced
robots and Internet of ings systems’ (13 July 2017)
market/en/news/workshop-liability-area-autonomous-systems-and-advanced-robots-and-
internet-things-systems> accessed 5 November 2018.
80 ibid.
81 de Bruin (n 61) 486.
82 Case C-503/13 Boston Scientic Medizintechnik GmbH v AOK Sachsen-Anhalt Die Gesundheitskasse
OJ C 138/9.
83 Zipp (n 72) 165.
84 ibid.
85 See most famously Soule v General Motors 882 P 2d 298 (Cal 1994).
86 PLD art 6(1); ACL s 9(1).
Autonomy and the Law of the Faster Horse 17
jurisdictions use the consumer expectation test for determining defectiveness. is
test establishes defectiveness using an objective standard, based on the general
public’s legitimate expectations of the product’s safety.87
As noted earlier in Part III, the issue of potential defects in the AVs physical
components (or their component parts) is straightforward, and no dierent to the
existing body of case law in the area.88 e question in relation to defects in the
AV’s programming and algorithms is the key legal issue. In a hypothetical scenario
where a plainti travelling in an AV suers harm due to the way the vehicle’s
programming responds to the situation, the consumer expectations test may create
a negative result for all stakeholders except the plainti in that situation. Imposing
liability based on the consumer expectations test in the design of autonomous
systems would likely create a result unpalatable to manufacturers and undermine
the incentive for the products to become widely available.
A consumer expectations framework is dicult in a product market where the
appropriate level of safety is dicult to understand for the general public or based
on erroneous expectations. e likely scenario in Irish and Australian Courts based
on existing case law is that the public expectation in relation to AVs is that they
are revolutionary machines free from human error and thus crash proof. As such,
regardless of the fact no autonomous system can be programmed to contemplate
or react to every situation presented to it appropriately, the public expectation
in relation to AVs is one of infallibility. e outcome in this type of scenario is
already clear. While the case of A v United Kingdom related to the supply of tainted
blood products, the Court found that consumer expectations of the risks posed
by blood products were ‘not known and accepted by society’.89 us, any problem
with the blood, regardless of the fact it is impossible for suppliers to screen for it,
enlivens strict product liability law. is places AV manufacturers in the dicult
situation of trying to ensure consumers understand the limits and risks of AVs,
whilst attempting to market them as vastly more safe than human operated vehicles
(which in turn creates a safer environment for all road users). With manufacturers
appearing to talk ‘out of both sides of the mouth’, this could ultimately create
signicant confusion among consumers and the courts.90
is model has long been rejected in the majority of United States jurisdictions.91
e courts there have recognised that in relation to complex products, the general
87 A v National Blood Authority [2001] 3 All ER 289; Glendale Chemical Products Pty Ltd v
Australian Competition and Consumer Commission (1998) 90 FCR 40.
88 KC Webb, ‘Products Liability and Autonomous Vehicles: Who’s Driving Whom?’ (2016) 23(4)
Richmond Journal of Law & Technology 9, 30–31.
89 A v National Blood Authority (n 87) [32], [55].
90 Webb (n 88), 29-30.
91 See Soule v General Motors, 882 P 2d 298 (Cal 1994); Ortho Pharmaceutical v Heath 722 P2d
410 (Colo 1986); Heaton v Ford Motor Company 435 P2d 806 (Or 1967); Cf Campbell v General
Motors 32 Cal3d 112 (Cal 1982); Barker v Lull Engineering 573 P2d 443 (Cal 1978); American
Law Institute, Restatement (ird) of Torts: Products Liability (1997), §402A, under which strict
liability only applies to manufacturing defects (not design defects or defects in information
provided with the product).
18  
public typically has no idea how safe a product is in any given situation, and that
relying on consumer expectations creates opacity and uncertainty on how the law
is to apply in any situation. Two California Supreme Court cases highlight the
underlying issue with relying on consumer expectations in technically complex
products. In the Campbell v General Motors decision, the Court held that a bus
lacking handle grips for standing passengers could be determined to be defective
based on consumer expectations of what constitutes a safe bus for standing
passengers.92 Just over a decade later, the same Court in Soule v General Motors
was forced to consider a case where a high-speed crash caused a front wheel to
travel backwards into the driver’s footwell and cause injury.93 e decision at rst
instance was based on the consumer expectations test, and found for the plainti,
who argued that the car’s design should have prevented the wheel from leaving its
frame and bracket. On appeal the California Supreme Court rejected this analysis,
arguing the ordinary consumer lacks the knowledge about engineering principles
and vehicle design to determine if the wheel should have moved from its position
in an accident at that speed.
e Court instead relied on conducting a risk/utility analysis where the Court
considered a number of factors to determine if the design was defective, including
the relative cost of the product (balancing the need for high levels of consumer
protection, but ensuring products are aordable); the gravity of the potential harm
from the claimed defect; and the cost and feasibility of eliminating or minimising
the risk.94
As the underlying technology, capabilities, and limitations of AVs are not likely
to be well understood by the ordinary consumer, it appears likely the existing
Australian and EU jurisprudence on defectiveness will be interpreted to capture
harms caused by all decisions taken by autonomous systems. Such an absolutist
application of the consumer expectations test ignores the general public benet in
vastly reducing the overall number of accidents, 90 per cent of which are caused by
human error. e consumer expectation test cannot evaluate the comparative costs
and benets of a technically complex but innovative product or design.
(a) Warnings, representations and disclaimers
is analysis can also be applied to products that are defective due to inadequate
warnings. Given their size, weight, and ability to travel at high speeds, automobiles
have been recognised as inherently quite dangerous for over a century now.95 is,
92 Campbell v General Motors (n 91).
93 Soule v General Motors (n 91).
94 See for example Seattle-First National Bank v Tabert 542 P2d 774 (WA 1975).
95 See for example MacPherson v Buick Motors 217 NY 382 (NY 1916); Henningsen v Bloomeld
Motors 161 A 2d 69 (NJ 1960).
Autonomy and the Law of the Faster Horse 19
along with the inherent complexity of the underlying autonomous technology,
makes it likely a manufacturer in both Australia or Ireland will owe a duty to
consumers to warn about the dangers of the product.96 However, the inherent
complexity of the driving decision making processes in an AV make it dicult
to adequately convey the nature of potential risks in a practical manner.97 Courts
would likely look to a minimum warning requirement based on codes of conduct,
or standard setting organisations, rather than rely on a more nebulous standard of
what an individual consumer expects in terms of warnings.98
e warnings, advertising, and other related conduct by manufacturers may also
raise concerns. In the EU, the prohibition on misleading information derives
from the Unfair Commercial Practices Directive.99 Article 6(1) of the Directive
denes misleading actions as those containing information that is either false or
deceives the average consumer. Information can be deceptive even when factually
correct, if the information is misleading. Most pertinently for the AV context, this
includes the product’s main characteristics such as the risks, tness for purpose,
and the results of tests carried out on the product.100 is also applies to material
information that is omitted ‘if, in its factual context, taking account of all its
features and circumstances and the limitations of the communication medium’,
the omission undermines the consumer’s ability to make an informed decision.101
Sections 42 through 46 of the Consumer Protection Act 2007 in Ireland enshrine
these articles of the Directive into Irish law. e ACL contains broadly similar
provisions, namely a general prohibition on misleading or deceptive conduct, and
a specic prohibition on false or misleading representations.102 In the past, this has
included claims made by vehicle manufacturers regarding the nature and quality of
specic features.103
Consider for example the oen-bold claims of Tesla founder and eternal AV
evangelist Elon Musk. His claims in relation to the capacity of Teslas to drive in
AutoPilot mode safely are problematic aer several fatalities involving Tesla cars.
ese include the vehicle mistaking the back of a truck for the sky or a driver
reportedly watching a Harry Potter lm when the AutoPilot mode was engaged.104
Manufacturers will need to be wary of making bold claims about the revolutionary
96 Chappel v Hart (1998) 195 CLR 232; Rosenberg v Percival (2001) 205 CLR 434.
97 Geistfeld (n 58) 1621.
98 Worsley v Tambrands Ltd [1999] EWHC 273.
99 European Parliament and Council Directive 2005/29/EC of 11 May 2005 concerning Unfair
Business to Consumer Commercial Practices in the Internal Market [2005] OJ L149/22.
100 ibid art 6(1)(b).
101 ibid art 7(1).
102 ACL ss 19, 29(1)(a).
103 Australian Competition and Consumer Commission v Audi Australia Pty Ltd [2007] FCA 1990.
is case involved misleading advertisements where a vehicle with a 7-seat conguration was
available at a certain price, but in fact the 7-seat conguration was an optional extra cost.
104 Damien A Riehl, ‘Car Minus Driver: Autonomous Vehicles Driving Regulation, Liability, and
Policy’ (2018) 35(5) e Computer and Internet Lawyer 1, 12.
20  
nature of AVs, given the fact many of these claims will be based on testing data
rather than real world experience.105
Further, the o-mooted business model for AVs involves end customers hailing a
vehicle for an individual trip or having a subscription to an AV service to use the
vehicles as needed. As this author and others have noted, technology companies
already active in the transportation sector (such as Uber) are keen to disclaim all
liability under product liability law. is is despite the fact that producers cannot
contract out of either European or Australian product liability laws.106 Central to
this line of reasoning is that customers using these applications are using a service,
with the intermediary providing only a technology platform that connects drivers
and those in need of transport. e Australian Competition and Consumer
Commission has in the past engaged in investigations regarding Uber’s terms of
service, alleging they contain unfair terms and are unconscionable as a result of
their attempts to place all liability on drivers.107 Across both Ireland and Australia,
the issue of unfair contracts has become an increasing concern for regulators and
courts.108 By the time AVs become commonplace, it is likely new reforms will exist
to better protect consumers from such unfair terms.109
(b) State of the art technologies
Given the nascent state of AV technology, a controversial exception to strict
product liability found in both the Irish and Australian context is relevant.
Referred to as the development risk defence, the test was adopted at the urging of
the UK and Ireland as part of the PLD to ensure new and innovative goods could
more easily come to market.110 Article 7(e) of the PLD states that manufacturers
cannot be liable for defects when the ‘state of scientic and technical knowledge
105 See Hawkins (n 5); PlainSite (n 30).
106 See for example Alice Witt, Nicolas Suzor and Patrik Wikström, ‘Regulating Ride-Sharing in the
Peer Economy’ (2015) 1 Communication Research and Practice 174; Christopher Koopman,
Matthew Mitchell and Adam ierer, ‘e Sharing Economy and Consumer Protection
Regulation: e Case for Policy Change’ (2015) 8 e Journal of Business, Entrepreneurship &
the Law 529; Matthew Sinclair, ‘Fair and Ecient Regulation of the Sharing Economy’ (2016)
36 Economic Aairs 204; Connor Hogg ‘e Sharing Economy: e Modern Consumer’s
Challenge’ (2018) 26(4) Australian Journal of Competition and Consumer Law 247.
107 Tom Westbrook, ‘”A lot to look at there”: ACCC chief ags Uber Eats probe’ (e Sydney
Morning Herald, 23 April 2018)
at-there-accc-chief-ags-uber-eats-probe-20180423-p4zb4b.html> accessed 5 April 2019.
108 See for example e Governor and Company of the Bank of Ireland v Patrick McMahon and Angela
McMahon [2018] IEHC 455; Ulster Bank Ireland Limited v Liam Costelloe and Gabrielle Bishop
Costelloe [2018] IEHC 289; Australian Competition and Consumer Commission, ‘ACCC
Enforcement Priorities 2019’ ps://www.accc.gov.au/about-us/australian-competition-
consumer-commission/compliance-enforcement-policy-priorities> accessed 1 February 2019.
109 Australian Consumer Law Review, Final Report (Consumer Aairs Australia and New Zealand,
March 2017) /uploads/sites/86/2017/04/ACL_Review_Final_
Report.pdf?_sm_au_=iVVPM H0Z008DsW8v> accessed 22 February 2019.
110 de Bruin (n 63) 490.
Autonomy and the Law of the Faster Horse 21
at the time when he put the product into circulation was not such as to enable the
existence of the defect to be discovered’.
Australia adopted the test soon aer, citing the same innovation policy ground as
well as a need to ensure Australian law was harmonised with the EU to ease import
and export barriers.111 Section 142(c) of the ACL is draed with almost exactly the
same wording to that found in article 7(e) of the PLD quoted above. e provisions
are also similar in that there is a dearth of precedent that interprets the defence.
However, from the albeit limited case law of two decisions, Australian courts have
been more willing to accept the applicability of the test compared to their European
counterparts. e dierence in interpretation could have signicant eects on any
legal cases brought in relation to an allegedly defective AV. is is because the
divergence relates to instances where the potential outcome of the defect is known,
but there are no scientic or technical means of discovering it during production.
e Court of Justice of the European Union and the ueen’s Bench in the UK have
both decided that the producers cannot avail of the development risk defence as
soon as they know about the risk, even if there is no way of testing that can identify
the risk.112 ese decisions took place in relation to contaminated blood, where at
the time, the state of the art was such that the suppliers were unable to detect the
specic diseases in the blood.113
In comparison, the o-cited Australian case of Graham Barclay Oysters was the
rst-albeit briey-to consider the applicability of the defence in Australia.114
Here the matter related to oysters contaminated with Hepatitis A, a risk that
was known at the time of production. However, the test to detect Hepatitis in
oysters was in its infancy at the time, was very costly, produced false negatives, and
destroyed the oyster in question so that only representative samples could be tested.
Unlike the Courts in Europe, the Australian Federal Court (and the Full Court
on appeal) agreed that producers could avail of the state of the art defence in the
circumstances.115 A decade later the Federal Court (and on appeal) armed that
the defence was available if the risk was known about by producers but the state
of the art meant it was unreasonable to think the defect could be detected during
production.116 e Court accepted that grounds such as practicality and industry
norms were to be considered, completely antithetical to the test in Europe.
111 See for example Australian Law Reform Commission (ALRC), Product Liability Report No 51
(1989), 47–49.
112 European Commission v United Kingdom [1997] 3 CMLR 923; A v National Blood Authority
(n88).
113 e decision by the CJEU concerned specically an action by the Commission against the UK for
their failure to properly transpose art 7(e) PLD.
114 Ryan v Great Lakes Council [1999] FCA 177; Ryan v Graham Barclay Oysters Pty Ltd; Great
Lakes Council v Ryan; New South Wales v Ryan [2000] FCA 1099.
115 ibid.
116 Peterson v Merck Sharpe & Dohme (Australia) Pty Ltd [2010] FCA 180; Merck v Sharp and
Dohme (Australia) Pty Ltd v Peterson [2011] FCAFC 128.
22  
Applying these tests to the AV context, it seems almost certain that manufacturers
will be unable to rely on any development risk defence in European jurisdictions
such as Ireland. e exceedingly high bar set out in the case law suggests that the
defence will be unavailable as soon as the risk of the defect is known about, even
if manufacturers are incapable of testing to identify every instance where the
autonomous soware controlling the vehicle may not work. AVs and contaminated
blood are similar: both are products where the public has an expectation of
complete safety, when in actuality, the nature of the products means that it would
be simply too costly, too time consuming, and too impractical to identify rare
defects during production. Conversely, these realities mean that Australian courts
are likely to accept manufacturers relying on this defence, as the value of having
such an innovative and benecial product on the market vastly outweighs the slight
risk that not all errors in the AV’s decision-making algorithms are identied and
xed prior to release.
A codicil to the issue of how innovative technologies interact with product
liability law relates to the oen limited lifespan of support for electronic goods.
Over time, internet connected consumer technology products, such as computers
and smartphones, require updates to close vulnerabilities and ensure the product
remains otherwise up to date. ere is considerable debate as to how long a
company ought to provide updates and other support to users of aging products.117
For example, Microso oered updates and support for the incredibly popular
2002 operating system Windows XP until 2014. In comparison, Apple has faced
recent action from the UK Competition and Markets Authority for its dubious
practice of slowing the performance of older model iPhones aer launching a new
operating system for their phones.118
More and more internet connected devices enter the market as part of the Internet
of ings boom, with the number expected to reach 29 billion during 2022.119
is raises the question of whether a product is defective if the manufacturer
does not provide adequate updates that cause the product to stop functioning at
the level a consumer is entitled to expect (such as exposing the user to security
vulnerabilities). In the AV sector, this question is particularly pertinent in the
event that a manufacturer becomes aware of an error in their vehicles’ decision-
making algorithms or the like, and fails to update the product. In the United
States, the Federal Communications Commission has been placing pressure on
manufacturers to be clearer with consumers about the frequency and importance
117 Riehl (n 104) 12.
118 Competition and Markets Authority, ‘Apple pledges clearer information on iPhone performance’
(United Kingdom Government, 22 May 2019) uk/government/news/apple-
pledges-clearer-information-on-iphone-performance> accessed 26 May 2019.
119 Ericsson, ‘Internet of ings Forecast’ y-facts/
ericsson-worldwide/india/authored-articles/ushering-in-a-better-connected-future> accessed 21
April 2022.
Autonomy and the Law of the Faster Horse 23
of their security updates.120 e obligation on AV manufacturers will be far greater
given the comparative cost of the product and the danger of failing to provide
adequate upgrades for the vehicles. Ultimately, regulators and courts will be forced
to consider when a lack of update amounts to a product defect and how long
manufacturers are required to provide support for older vehicles.
IV. A New Regime for Autonomous Vehicles
Part III of this paper demonstrated that while existing liability structures are
equipped to deal with many of the issues raised by AVs, further thought regarding
how to better determine and apportion liability in the context of incidents
involving AVs is necessary. Vehicles with SAE Levels 3 – 5 autonomy are likely to
be involved in a far fewer number of incidents, but the cost of damage incurred
in an incident is likely to increase given the complexity of the technologies and
components within the vehicle.121 As noted throughout this paper, the complexity
and nature of AV technology raises dicult questions about how to best determine
and apportion liability in the future. Far more discussion is necessary in both
Australia and Ireland about the unique liability issues for AVs to ensure consumers,
vehicle producers, AV technology developers, and regulators are condent in the
regulatory framework.122 is nal part of the paper will examine the current
attempts to regulate AVs and assess the ecacy of proposed liability reforms.
IV. I . Preliminary eorts
(a) International experiences
As the US State home to the largest technology companies, California has appeared
the most active and interventionist of the present regulatory eorts regarding AVs.
Phoenix, Arizona (and to a lesser extent Pittsburgh, Pennsylvania) have also been
the sites of major AV testing, with local and state governments in both jurisdictions
working to entice AV research investment through comparatively light touch
regulation.123 California and Arizona are the most active testing jurisdictions and
serve as a useful comparison to the two dominant regulatory approaches.
120 Dawn Chmielewski and Ina Fried, ‘US Regulators Want to Know Why Your Phone Isn’t Getting
Security Updates Faster’ (Recode, 9 May 2016) .vox.com/2016/5/9/11642640/
c-fcc-mobile-device-security> accessed 3 February 2019.
121 Lynne McChristian and Richard Corbett, ‘Regulatory Issues Related to Autonomous Vehicles’
(2016) 35(7) Journal of Insurance Regulation 1, 8.
122 ibid 9.
123 Roe (n 11) 326.
24  
Arizona’s regulations require only that AVs being used carry liability insurance and
that any supervising human driver have a valid license.124 While this minimalist
set of requirements was an eective enticement to draw AV trials away from
neighbouring California, the executive order could be criticised as allowing an
unacceptable level of risk-taking by companies. is is especially the case given the
low level of public buy-in to the policy, and complete dearth of expert engagement
or oversight.125 e fatal collision involving an Uber AV and a pedestrian crossing
a road whilst wheeling their bike exemplies the problem. In the lead up to the
incident, Uber removed the second testing driver from the car tasked with
monitoring safety, had no real-time means of monitoring driver distraction,
and the managers overseeing testing were responsible for too many drivers to
suciently audit behaviour. e incident rightly sparked community outcry, who
felt their safety had been sacriced in order to attract investment to the state.126 e
long-term goals of AV technology are hard to achieve when such a lackadaisical
approach to regulation is permitted.
Comparatively, California’s current regulations, adopted in February 2018, set out
a number of specic design and technology requirements that must be met before
a Manufacturer’s Testing Permit will be issued.127 Further, the regulations have
requirements regarding documenting and reporting of general safety outcomes, a
log of incidents, and supervising driver training procedures.128 e regulations also
provide for testing of SAE Level 5 vehicles without a driver, imposing even more
stringent obligations.129 California’s approach can be contrasted with Arizona’s
to demonstrate the necessity to strike a more even balance between protecting
public safety and encouraging innovation in the sector.130 e marked decrease in
testing in Arizona,131 compared to the steady rate of testing in California is a strong
indicator that a more robust regulatory regime with sucient mechanisms in place
can still encourage appropriate levels of risk-taking, innovation, and greater public
support for AV technology.
124 Arizona Revised Statutes, Title 28, Chapter 32, Article 1.
125 Mark Harris, ‘Arizona governor and Uber kept self-driving program secret, emails reveal’ (e
Guardian, 28 March 2018) .theguardian.com/technology/2018/mar/28/uber-
arizona-secret-self-driving-program-governor-doug-ducey> accessed 11 December 2018.
126 Johana Bhuiyan, ‘Months aer a self-driving crash, Uber is being slower and more deliberate as
it gets back on public roads to test its technology’ (Recode, 24 July 2018)
net/2018/7/24/17608434/uber-self-driving-crash-pittsburgh-arizona-testing-safety> accessed
11 December 2018.
127 California Code of Regulations, Title 13, § 227.
128 ibid § 227.38.
129 ibid.
130 Roe (n 11) 327.
131 Arian Marshall, ‘Aer the Fatal Crash, Uber Revamps its Robo-Car Testing’ (WIRED, 24 August
2018) testing> accessed 11
December 2018.
Autonomy and the Law of the Faster Horse 25
(b) e Australian approach
In Australia, the National Transport Commission (‘NTC’) is the independent
statutory body tasked with investigating road transport issues and law reform
proposals.132 e NTC has taken an active interest in the AV issue since 2015.
While no concrete policy exists as of yet, the general consensus forming at the NTC
is that gradual but signicant law reform will be necessary to incorporate AVs into
the Australian transportation ecosystem.133 Broadly, the NTC’s current initiatives
can be divided into work on ensuring road rules and related laws are equipped to
accommodate AVs, and law reform to modernise road accident insurance regimes.
In November 2016, the NTC was directed to develop a series of legislative reform
options to better clarify how laws governing drivers and the act of driving apply
to AVs. is was envisaged to eventually consider the legal obligations that should
govern AVs operating in the future.134 roughout 2017 and 2018, the NTC
engaged stakeholders and published a policy paper in May 2019, which identied a
range of issues in need of reform, including:
• e need to alter existing laws to redene what constitutes a driver and
who is legally responsible for an AV on the road;
Creating a nationally consistent framework to ensure AVs are commercially
viable;
Ensuring the law does not attempt to force new technologies into existing
human centric laws; and
Recognising the current uncertainty globally about the development of
AV technology.135
While transport ministers from across Australia have agreed to work to create a
uniform approach, presently a number of jurisdictions within Australia have
created a set of limited permits available for those wishing to operate AV trials.136
e policy discourse relating to liability is in a far more embryonic state. e NTC
is currently engaging with stakeholders to consider whether existing insurance and
132 National Transport Commission Act 2003 (Australia).
133 Araz Taeihagh and Hazel Si Min Lim, ‘Governing Autonomous Vehicles: Emerging Responses
for Safety, Liability, Privacy, Cybersecurity, and Industry Risks’ (2019) 39(1) Transport Reviews
103, 111.
134 National Transport Commission, ‘Changing driving laws to support automated Vehicles Policy
paper May 2018’.au/sites/default/les/assets/les/NTC%20Policy%20
Paper%20-%20Changing%20driving%20laws%20to%20support%20automated%20vehicles.
pdf> accessed 22 May 2019.
135 ibid.
136 VicRoads, ‘Automated Driving System (ADS) Permit Scheme’ (28 September 2018)
www.vicroads.vic.gov.au/safety-and-road-rules/vehicle-safety/automated-and-connected-
vehicles/testing-of-automated-vehicles> accessed 22 May 2018; Transport for NSW, ‘Regional
Automated Vehicle Trials’ ransport.nsw.gov.au/projects/programs/smart-
innovation-centre/regional-automated-vehicle-trials> accessed 22 May 2018.
26  
liability frameworks are equipped to incorporate AV accidents, or whether a new
insurance scheme will need to exist. As of yet there is no consensus on how AV
liability will be addressed in Australia.137
(c) e Irish approach
e west coast of Ireland is famous for challenging driving conditions. In 2018,
some manufacturers such as Jaguar Land Rover stated that it would be the ultimate
proving ground for AVs as ‘we have grass on the road, we have sheep on the road,
we have stones on the road, we have potholes and we have very interesting weather
conditions’.138
In October 2021, the Irish Minister for Transport introduced the Road Trac and
Roads Bill 2021 to the Dáil Éireann. As at late May 2022, the Bill remains at the
Second Reading stage. e AV related content in the Bill is minimal, proposing
only to introduce a denition of AV,139 and expand the denition of driving to
include periods of time where an AV is operating the vehicle.
e Bill’s Explanatory Memorandum notes that these changes are not intended to
be the entirety of domestic law on this issue, but merely to facilitate the testing of
AVs on Irish roads.140 In the meantime the Bill’s Explanatory Memorandum notes
that the Oireachtas and Government will continue to ‘observe developments in the
eld before it will be clear what legislation is needed’.141
IV.II. Proposed regulatory frameworks for autonomous vehicles
(a) Driver responsibility: e surgical robotic experience
e uptake of robotics in the medical surgery context has been observed by
Hubbard and others as a technology analogous to AVs and only a few years further
into their process towards widespread adoption.142 Surgical robots are deployed
to improve the precision of certain aspects of procedures. Much like drivers in the
137 National Transport Commission, ‘Motor Accident Injury Insurance and Automated Vehicles
u/current-projects/motor-accident-injury-insurance-and-automated-
vehicles/?topicId=1306> accessed 22 May 2018.
138 Colm Gorey, ‘We Want to Make Ireland the Centre of the Universe for Autonomous Cars’
(Silicon Republic, 29 June 2018) -
rover-john-cormican> accessed 20 November 2018.
139 Road Trac and Roads Bill 2021 (Ireland) s 5(a)(i): ‘“autonomous vehicle” means a mechanically
propelled vehicle that has been designed, constructed, adapted or modied to move autonomously
for certain periods of time but in respect of which driver intervention is still expected or required”.
140 Road Trac and Roads Bill 2021 Explanatory Memorandum, 4.
141 ibid.
142 See for example F Patrick Hubbard, ‘Sophisticated Robots: Balancing Liability, Regulation, and
Innovation’ (2014) 66 Florida Law Review 1803.
Autonomy and the Law of the Faster Horse 27
AV context at SAE Levels 3 or 4, surgical robots take control for certain aspects of
the procedure, with the surgeon able to take control to mitigate harm if an error
occurs.143 Given the complexity of both technologies, the diculties in proving
and apportioning liability, Hubbard suggests similar legal principles should be
applicable in both technologies.144
is analysis is developed by Schroll, who argues that consumers, like surgeons,
should be liable as they have chosen to purchase an AV, and thus voluntarily assumed
the risk relating to the product. She suggests that those using AVs ‘understand
that there is a risk of the technology malfunctioning and causing an accident, but
they decide to take on that risk because the benet of using the car outweighs the
risk’.145 e argument is also couched in terms of practicality, noting liability claims
relating to an accident are far easier to coordinate between the individuals involved
instead of claims against a large and likely multinational manufacturer. Schroll
contends that the ability to meet in person and avoid navigating complex corporate
structures makes the case for individual driver responsibility.146
However, these analyses do not consider the vastly dierent contextual settings for
surgical robots and AVs. e market of AV users is the same market as those who
currently own vehicles, who for the most part meet only a minimum age and skill
qualication (a driving test). Surgeons on the other hand have years of base training
and continuous on the job training. Concerns that drivers will be incentivised
to behave recklessly without the risk of being forced to pay damages is no longer
applicable. is is because it is envisaged by AV manufacturers that drivers will
have no need and eventually no ability to intervene in the operation of the vehicle.
Surgeons only use robotics as an assisting technology, with greater knowledge of
when to intervene safely compared to the ordinary driver, and more time to do so
as compared to split second decision making in the AV context.147
Further, these arguments ignore the role economic incentives play in decision
making. e widespread deployment of AVs is likely to dramatically decrease
road casualties, decrease the need to own personal transport, and benet the
environment by vastly reducing the number of vehicles needed in a location. It is in
the interest of individuals and society at large to adopt AVs given these benets. As
such, a liability structure where the person in the vehicle is responsible regardless
of the cause of the accident or their inability to mitigate harm incentivises people
to avoid using AVs for fear of being liable.148 Liability frameworks must strike a
balance between adequate protection of individuals and encouraging innovation
143 Roe (n 11) 328-29.
144 Hubbard (n 142).
145 Schroll (n 31), 815.
146 ibid 816.
147 Roe (n 11) 336.
148 Dana M Mele, ‘e uasi-Autonomous Car as an Assistive Device for Blind Drivers: Overcoming
Liability and Regulatory Barriers’ (2013) 28 Syracuse Journal of Science and Technology Law 26,
27.
28  
and uptake in new technologies and products.149 Assigning liability solely to users of
AVs will create insucient demand for such vehicles, and ignores product liability
principles that sees traditional manufacturers as those best placed to mitigate risks.
It is these dierences that raise unique liability questions in the AV context, as there
is very little or no capacity for human intervention. At present, the approach to
determining and apportioning liability in the surgical robotics context has been
ex-post, with courts engaging in analysis to determine whether the manufacturer,
hospital, or surgeon is responsible in the specic circumstances.150 is approach is
unpalatable given the number of vehicles on the road compared to the number of
surgeries performed. An ex-ante approach is necessary.
(b) Intermediary liability
e high cost of manufacture and the fact that the slack capacity of an AV can be
eciently utilised by others in need of transport have created an industry consensus
that vehicles will most likely be hired as needed rather than owned by consumers.151
As such, intermediaries (or the manufacturer oering access to the vehicles as
a subscription) will likely become important stakeholders in the market and the
liability issues that will arise.
Both the PLD and the ACL contain provisions ensuring that intermediary
suppliers of goods may be liable for defects. is is borne of the principle that the
end consumer should not be the party to bear the burden of tracing and bringing
suit against the manufacturer in the oen-complex modern market economy.
Intermediaries covered by these provisions include importers, wholesalers, and
retailers.152
e proposition makes intuitive sense in the AV framework. As large-scale
purchasers of AVs, car-sharing intermediary companies will have greater capacity
to pressure manufacturers through their signicant purchasing power compared to
individual consumers. erefore, placing a liability burden on them will ‘encourage
competition among manufacturers to improve the safety of the cars they make for
companies’.153 Further, as the intermediaries are likely to be the actors responsible
149 ibid.
150 See for example Mracek v Bryn Mawr Hospital 363 Federal Appendix 925, 926 (3rd Cir 2010);
Brown v Grin 505 SW 3d 777, 778 (KY 2016); Taylor v Intuitive Surgical Incorporated 389 P
3d 517, 526-28 (WA, 2017). Both Ireland and Australia have recently acquired their rst surgical
robotic tools, the same ones used in these cases. See Pat Harrold, ‘e €2.6m surgical robot
in Limerick that’s not allowed to fully operate’ e Irish Times (Dublin, 26 November 2018)
rgical-robot-in-limerick-
that-s-not-allowed-to-fully-operate-1.3705558> accessed 22 March 2019.
151 Schroll (n 31) 801.
152 ibid. See for example Lambert v Lewis [1982] AC 225; Fisher v Harrods [1966] 1 Lloyd’s Rep
500; Devilez v Boots Pure Drug Company Ltd (1962) 106 SJ 552.
153 Schroll (n 31) 818.
Autonomy and the Law of the Faster Horse 29
for placing large numbers of AVs on the road, their ability to prot from this
conduct could justify placing liability upon them, the basis for strict liability under
the PLD and the ACL.
Schroll suggests that the intermediary companies would be able to avoid liability
in the US by relying on the Graves Amendment.154 is provision shields car rental
businesses from liability in relation to harms caused by negligent consumers that
hire their vehicles. is justication is not applicable to the Irish or Australian
context as no such provisions exist in either jurisdiction. Further, Schroll’s argument
conates negligent driving on the part of the hirer with a manufacturer creating a
defective product. A hiring customer cannot negligently operate the vehicle when
they are not driving it. A long line of case law suggests that customers can bring
a product liability claim against a rental or hire company as they are part of the
supply chain and were the party to directly engage with the consumer rather than
the manufacturer in these situations.155
e case for end users of AVs being able to sue intermediaries under the PLD or
ACL is compelling and clear, as it enables consumers to more easily make a claim
under product liability law. However, as the following sections will elucidate, the
ecacy of traditional product liability law in the realm of AVs is questionable based
on broader public policy considerations.156 Instead, an approach that departs from
traditional principles of strict liability and the current denition of defectiveness
appears more appropriate.
(c) Programmer and manufacturer liability
For those riding inside a vehicle operating autonomously at SAE Levels 3–5, there
is no convincing causative or policy argument for them to be held liable under any
tort or product liability framework. A defective product claim in relation to an
AV would centre on either an error in the vehicle’s programming, or one of the
physical components of the vehicle – party claimants would naturally look to the
manufacturer in this instance.
ose creating autonomous technologies must be deeply thoughtful about the way
in which they program their inventions. To this day, Asimov’s famed ree Laws of
Robotics remain the most popular understanding of the seemingly inherent duty of
154 49 US Code§ 30106.
155 See Grant v Australian Knitting Mills [1936] AC 85; Martin v orn Lighting Industries Pty Ltd
[1978] WAR 10; Tarling v Nobel [1966] Arg LR 189; Fletcher v Toppers Drinks Pty Ltd [1981]
2 NSWLR 911; MacLachlan v Frank’s Rental (1979) 10 CCLT 306; omas v Foreshore Marine
Exhaust Systems Pty Ltd [2005] NSWCA 451; Suthern v Unilever Australia Ltd [2007] ACTSC
81.
156 Mele (n 148) 42.
30  
autonomous technologies and their creators to protect the general public.157 ey
are set out as follows:158
1. A robot may not injure a human being or, through inaction, allow a human
being to come to harm.
2. A robot must obey the orders given to it by human beings, except where
such orders would conict with the First Law.
3. A robot must protect its own existence as long as such 3. protection does
not conict with the First or Second Law.
Product liability principles would dictate that those programming and
manufacturing AVs have a similar obligation to design a set of fundamental
principles to guide the decision-making and operations of the vehicle. ese would
likely revolve around ensuring the AV complies with road rules and maximises
human safety, evaluating a ‘driving event’ to comply with a series of coded
instructions.159 is makes intuitive sense, as then any incidents that are the result
of the coding can be deemed a defect and thus within the scope of the programmer/
manufacturer’s product liability obligations.
While this argument is appealing for both its simplicity and the ease in which it
can bring AVs into the fold of existing product liability law, it fails to consider that
autonomy may upend existing conceptions of liability. e next section of the paper
considers the case for a no-fault liability structure in the AV context. It argues that
the potential positive externalities the deployment of AVs oers and the nature of
autonomy may justify the costs of operating such a scheme.
(d) A no fault liability scheme
For those that suer harm as a result of an incident involving an AV, their ‘interests
do not change due to the mere fact that the vehicle operated automatically’.160
However, as this paper has noted throughout, the calculus in determining how
liability is to be apportioned in the AV context is far from simple. At present,
almost all accidents are the result of human error rather than a vehicle defect. Even
the majority of defects are in fact the result of poor maintenance on the part of
the vehicle’s owner.161 AV developers are condent that they will be able to reduce
trac casualties by up to 90 per cent and remove the need for most or all human
157 MIT Technology Review, ‘Do We Need Asimov’s Laws?’ (16 May 2014) tps://www.
technologyreview.com/s/527336/do-we-need-asimovs-laws/> accessed 1 April 2019.
158 ibid.
159 Chris Urmson and others, ‘Autonomous Driving in Urban Environments: Boss and the Urban
Challenge’ (2008) 25 Journal of Field Robotics 425, 457.
160 Melinda Florina Lohmann, ‘Liability Issues Concerning Self-Driving Vehicles’ (2016) 7(2)
European Journal of Risk Regulation 335–338.
161 ibid 337.
Autonomy and the Law of the Faster Horse 31
control of vehicles.162 An ex-ante policy solution proposed by McChristian and
Corbett is that of a no fault liability framework.163 ey propose altering the design
of vehicle insurance regimes, so that compensation claims can be awarded without
the need to assign liability to any party/parties.
e proposal attempts to incentivise end users to use/hire and manufacturers
to produce AVs. e existing liability framework is cognisant that in the vehicle
context, human error is almost always the cause of accidents and that vehicle
owners are best positioned to take measures to reduce the risk of harm, such as
safe driving and adequate vehicle maintenance.164 As such, drivers must be insured
so that civil actions can be brought against them for harms they may cause. In an
incident involving a vehicle that was operating autonomously, a driver will not be
held liable for harm without a nexus of causation.
AVs benet society due to their social utility. Imputing liability for a defective
AV onto the passenger in the vehicle undermines traditional product liability
principles. It also would make passengers liable under a strict liability standard,
compared to drivers who are currently held to a negligence standard in tort.
is would not only be unjust, but would likely aect the sale of AVs, impeding
innovation and restoring high accident rates. For these reasons, driver-liability
should not be shied to the vehicle owner.165
Given that AVs are to operate with little to no human intervention, manufacturers
are those most capable of reducing the risk of harm by designing and producing
high quality vehicles without defects. Further, they are best placed to inform the
end user of any product risks and mitigation strategies. Relying on the existing
product liability framework to claim against manufacturers is not practicable
in relation to most incidents involving AVs. A claim against a manufacturer in
relation to product liability in both Australia and Ireland is far more complex,
time-consuming, and costly compared to an insurance claim against a driver. is is
especially the case in relation to technology products.166 ese factors make claims
extremely prohibitive, especially in relation to small claims for personal injuries
and/or property damage. Further, there is a concern that manufacturers will be
disincentivised to produce AVs if their liability burden is increased. is would be
a problematic outcome as AVs capable of dramatically decreasing vehicle accidents
would be highly socially desirable.167
162 Daniel J Fagnant and Kara M Kockelman, ‘Preparing a Nation for Autonomous Vehicles:
Opportunities, Barriers and Policy Recommendations’ (Eno Center for Transportation, October
2013) accessed 8 February 2019.
163 McChristian and Corbett (n 120) 9.
164 Lohmann (n 160) 338–39.
165 Sellwood (n 27), 855.
166 McChristian and Corbett (n 120) 9.
167 Lohmann (n 160) 338–39.
32  
Instead, law reform eorts must treat AVs (and potentially other autonomous
products) as requiring a solution that acknowledges their unique status. AVs
have the potential to be a highly valuable social good that enhances public safety,
benets the environment, and generally promotes the mobility of all citizens.168 As
such, reform eorts in relation to vehicle insurance and product liability law should
acknowledge the extremely technologically complex nature of AVs and establish
a (possibly government operated) no fault compensation scheme. e scheme
should operate based on a presumption that the autonomous technologies within
the vehicle were the cause of any incidents, unless the manufacturer can prove
otherwise. is would act as the incentive for manufacturers to create safe products
in much the same way as the risk of liability under product liability legislation.169
Over time, a greater share of vehicles will become autonomous and personal vehicle
ownership will decline as consumers acquire the services of an AV only when
they require transport. AVs will replace the need for individual vehicle insurance.
ough a government operated no fault liability scheme may be expensive to set
up or administrate, there are eciencies for all stakeholders by greatly reducing the
time and expense of processing and litigating insurance claims or product liability
litigation.170 e heavy reduction or near elimination in the social, environmental,
and economic costs incurred as a result of the current road transport paradigm
means that the introduction of AVs is likely to be a net positive for society.
While the practicalities of such a no-fault liability scheme are beyond the
scope of this paper, it will suce to say there are a range of methods that could
be used to fund such a scheme. e funds for such a scheme could be collected
from the money drivers currently pay towards car insurance as some form of tax.
Alternatively, it could be charged to manufacturers, who could then pass on and
spread the costs across AV purchases-much in the same way manufacturers spread
the cost of their strict liability burden in traditional product liability situations.171
Another approach is for a fee to be levied to manufacturers on the basis of
vehicles produced, and to consumers or AV eet operators based on the number
of trips taken. Schroll argues this approach maximises fairness as it ensures larger
manufacturers pay proportionate to the number of vehicles they have produced
that may be in an accident, and for eet operators and consumers to pay ‘based
on their frequency of use so that those most at risk of needing payouts will pay
the most in taxes’.172 ese are of course preliminary proposals. Over time, metrics
about the safest manufacturers and driving jurisdictions can be determined,
ensuring that the tax rates are fairly apportioned. is can only serve to further
incentivise manufacturers to produce the safest possible vehicles.
168 Viktória Ilková and Adrian Ilka, ‘Legal aspects of autonomous vehicles’ (21st International
Conference on Process Control, Štrbské Pleso, Slovakia, June 2017) 428; Lohmann (n 160) 335.
169 Lohmann (n 160) 339.
170 Lohmann (n 160) 339.
171 Ilková and Ilka (n 168) 431.
172 Schroll (n 31) 823.
Autonomy and the Law of the Faster Horse 33
ere are of course likely to be large barriers to establishing a new government
operated scheme-namely from those concerned about government overreach or
the unfairness of allowing corporations to obviate their liability regarding products
they designed.173 Further, competitors in the transportation industry (including,
for example, unions representing truck or taxi drivers) are likely to object, claiming
AVs will be unfairly advantaged in the market. Even those in the autonomy and
robotics industries may oppose such a move, arguing that ‘if only a particular
approach to development of [a specic autonomous technology] is favoured,
those pursuing other alternatives would be disadvantaged, thus potentially stiing
innovation the immunity is meant to promote’.174
However, no fault schemes already exist in a number of US states and in New
Zealand.175 In these circumstances, policymakers have determined it is appropriate
that there be a mechanism for the few who suer harm as a result of generally
benecial public goods to be compensated, without a need to prove fault. Irish
and Australian parliamentarians and governments have proposed similar no fault
liability programs for the small minority who suer harm from vaccines given their
overwhelming net positive societal impact.176 AVs have the potential to be similarly
benecial to the public at large, with few outliers suering harm. As such, a no-fault
liability scheme appears to be the most appropriate regulatory mechanism once AVs
are adopted. Without thoughtful consideration from regulators and legislators, the
damages incurred from AVs will fall within the ambit of legal regimes ill-equipped
to properly regulate them.177
For the EU at least, the case for such a no-fault framework has already received
a modicum of initial support. A 2017 European Union Committee on Legal
Aairs Report to the Commission on Civil Law Rules on Robotics argued that
sophisticated autonomous technologies and robots should be classied as a new
category of legal personality, with their own specic rights and obligations.178
Opinion was divided between a range of options spanning from maintaining
existing product liability laws, through to a general fund that would cover harms
173 Schroll (n 31) 830.
174 Hubbard (n 142) 1871.
175 Mark R Forwood, ‘Whither No-fault Schemes in Australia: Have We Closed the Care and
Compensation gap?’ (2018) 43 Alternative Law Journal 166; National Childhood Vaccine Injury
Act of 1986 (42 United States Code §§ 300aa-1-300aa-34).
176 ibid; Alan O’Keee, ‘Government Remains Committed To ‘No Fault’ Scheme to Help Victims
When Vaccinations Do Unintended Harme Irish Independent (Dublin, 17 February 2019)
-news/health/government-remains-committed-to-no-fault-
scheme-to-help-victims-when-vaccinations-do-unintended-harm-37823344.html> accessed 22
February 2019.
177 Daniel A Crane, Kyle D Logue, and Bryce C Pilz, ‘A Survey of Legal Issues Arising From e
Deployment of Autonomous and Connected Vehicles’ (2017) 23 Michigan Telecommunications
and Technology Law Review 191, 259.
178 Committee on Legal Aairs with recommendations to the Commission on Civil Law Rules on
Robotics, 27 January 2017 (2015/2103(INL)).
34  
caused by all autonomous systems and robots.179 ough the report is only a
series of hypothetical scenarios and mooted solutions, it substantiates the premise
outlined at the beginning of this paper. As Sellwood notes, ‘giving a legal status to
an AV is no longer considered science ction; instead it is modern reality’.180
For the time being, some manufacturers such as Google/Waymo, Mercedes-Benz
and Volvo have already pledged to accept legal responsibility in relation to their
AVs.181 However, this occurred in 2015. It remains to be seen if these manufacturers
will maintain this approach once AVs leave the testing phase and enter the market.
IV.III. Legal protection by design
e ability of AVs to make decisions completely independent from the person
travelling in the vehicle, and perhaps even the vehicle’s manufacturer, is an example
of a worrying trend for product liability regulators. e European Commission has
expressed concern that the current conceptions of what constitutes a product and
a defect has become impracticable in the wake of modern technology.182 Writing in
2018, the Commission succinctly summarised this general concern, stating:
2018 is not 1985. e EU and its rules on product safety have evolved,
as have the economy and technologies. Many products available today
have characteristics that were considered science ction in the 1980s. e
challenges we are facing now and even more acutely in the future – to name
but a few – relate to digitisation, the Internet of ings, articial intelligence
and cybersecurity.183
Autonomous products capable of acting and even learning without the control of
a manufacturer or human operator are one of these emerging concerns. AVs are
so technically complex that no ordinary consumer has an understanding of how
they function or what they can reasonably expect of them.184 Products that are
this complex and opaque to the end user create a risk that the ‘normativity that
is inherent in the computational infrastructure will overrule the normativity of
179 ibid [59].
180 Sellwood (n 27) 864.
181 Volvo, ‘Volvo Cars Responsible for the Actions of its Self-Driving Cars’ (20 October 2015)
accessed 18 October 2018; 60 Minutes, ‘Hands O the Wheel’
accessed 10 February 2022.
182 Commission, ‘Report from the Commission to the European Parliament, the Council and the
European Economic and Social Committee on the Application of the Council Directive on
the approximation of the laws, regulations, and administrative provisions of the Member States
concerning liability for defective products (85/374/EEC)’ (2018) COM 246 nal, 9.
183 ibid 1.
184 Mireille Hildebrandt and Laura Tielemans, ‘Data protection by design and technology neutral
law’ (2013) 29 Computer Law & Security Review 509, 516.
Autonomy and the Law of the Faster Horse 35
the written law’, leaving product liability laws incapable of providing adequate
protection.185
An emerging area of legal discourse has attempted to counter the concern that
technology is developing at a pace too rapid for law to ever keep up. Ambient law
theory instead proposes to enshrine legal norms into the underlying code and design
of new technology, rather than relying on written laws to inform manufacturers
of their legal obligations as they design products.186 Scholars such as Hildrebrandt
and Tielemans are optimistic this new way of conceptualising legal protections
will enable designers and manufacturers to compete to make the most consumer
friendly iterations of new technologies, as they vie to ‘provide creative solutions to
the requirements set by the legislator’.187
However, the ambient law discourse rst arose around the issue of data protection.
e issues AVs will be forced to consider are likely far more serious, given the
risk of harm in a vehicle collision. If product liability laws were encoded into
the underlying design of an AV, dicult ethical decisions quickly arise. Product
liability laws exist to bring about a general improvement in societal welfare by
incentivising the design of safer products and ensuring individuals are adequately
compensated in the event of harm suered due to a defective product. Deciding
how these general principles inform AV design will require determining acceptable
answers to complex moral, ethical, and legal questions such as the age old trolley
problem. is thought experiment places a tram on a track which is on a path set to
crush ve persons tied down to the track. A lever, however, is present which would
allow the tram to take an alternative route, where only one person is tied down to
the track. e experiment asks if one would change the tram’s direction, knowing
the tram would instead hit only one person on a dierent track, rather than ve.188
An AV placed in a similar situation may be forced to consider a situation where
it must decide whether to prioritise the welfare of its internal passengers and
collide with another road user, or instead focus on whatever outcome minimises
the economic and social cost of an accident.189 Unlike human drivers, AVs
have no survival instinct nor do they place more value on the passengers within
the vehicle as compared to other road users. is raises the question of whether
an autonomous product should be designed to act as the ‘least cost avoider’ and
act in a way to minimise harm suered, knowing that product liability law can
recompense those harmed.190 While in the abstract this may be the most palatable
185 Mireille Hildebrandt, ‘Legal Protection by Design: Objections and Refutations’ (2011) 5(2)
Legisprudence 223, 232–33.
186 ibid 223.
187 ibid 239-240.
188 Amar Kumar Moolayil, ‘e Modern Trolley Problem: Ethical and Economically-Sound Liability
Schemes for Autonomous Vehicles’ (2018) 9 Case Western Reserve Journal of Law, Technology
and the Internet 1, 10–11.
189 ibid.
190 ibid 11–12.
36  
solution, the traveller in any AV will in their own self-interest instead prefer the AV
to act to best protect them, even at the expense of other road users. e ambient
law considerations in relation to AVs are immense and challenging.
As Part IV of this paper has demonstrated, product liability law could be reformed
to better ensure those harmed by AVs are guaranteed the protections provided
under legal instruments such as the PLD or ACL. Nonetheless, manufacturers will
need to make dicult ethical and legal decisions about how the underlying design
of AVs complies with product liability law.
V. Conclusion
Regardless of the plethora of concerns raised in this paper, it is almost impossible
not to be enthralled by the prospect of AVs. e ability to eliminate the harms
suered on the road due to human error will likely save nations billions of euros a
year, not to mention countless lives. In addition to human welfare considerations,
AVs ‘have the potential to transform personal mobility and open doors to people
with disabilities, aging populations, and communities where car ownership is
prohibitively expensive’.191 Increased transport eciency will in turn bring about
environmental benets by drastically reducing the number of vehicles on the road.
Despite these considerable potential benets, consumers are naturally sensitive
to the risks posed by AVs. Previous incidents involving AVs have created a strong
emotional response, a natural human emotion against new technologies that can
cause harm, even if the technology drastically improves social welfare overall.192 In
an era of seemingly constant technological innovation, the law has oen failed to
keep pace with change.193 In the realm of AVs, the lack of legal certainty regarding
product liability has hampered the development of the industry, with technology
again outpacing the law.194
In summary, this author has argued that the existing conceptions of product and
defect are too narrow to eectively protect consumers using AVs. It was highlighted
that the autonomous decision-making technologies within AVs make it dicult
to determine what constitutes a defect. Product liability law was not draed to
consider the potential defectiveness of a decision taken by a product that can learn
and act independently, potentially requiring manufacturers ‘to conceive of every
possible situation the vehicles might encounter prior to their release’.195 ere is a
191 Ilková and Ilka (n 168) 428; Lohmann (n 160) 335.
192 Gary E Marchant and Rachel A Lindor, ‘e Coming Collision between Autonomous Vehicles
and the Liability System’ (2012) 52 Santa Clara Law Review 1321, 1335.
193 See for example Richard A Epstein, ‘Can Technological Innovation Survive Government
Regulation?’ (2013) 36 Harvard Journal of Law & Public Policy 87; Nathan Cortez, ‘Regulating
Disruptive Innovation’ (2014) 29 Berkeley Technology Law Journal 175.
194 Sellwood (n 27) 843.
195 Zipp (n 72) 163.
Autonomy and the Law of the Faster Horse 37
great deal of technical and legal work to be done to ensure that the PLD and ACL
are capable of identifying defects and proving causation.
Secondly, it was argued that the existing PLD and ACL are not designed to
consider products that greatly enhance social welfare at the rare expense of an
unlucky consumer. us, instead of focussing on policy solutions that have been
applied to existing products, it was contended that the approach oen adopted for
public health initiatives, such as vaccines, is a superior approach. It acknowledges
the broad social benets of the product vastly outweigh the individual harms
that can occur on occasion. As such, a no-fault liability scheme that centrally
compensates those that suer harm appears to be the best solution to encourage
innovation and adoption of the technology.
With the technology still in its infancy, there is time le for all stakeholders
to contemplate these issues. For new technologies, the law is too oen seen as a
lumbering nuisance that fails to keep pace with innovation.196 More proactive
research into the interaction between law and emerging technologies will enable
innovators and policy makers to make better decisions. is paper intended to
serve as an initial beachhead in the area of AVs and product liability law in two
jurisdictions not normally the focus of academic research on emerging technology
law and policy.
196 Se e for example Howard A Shelanski, ‘Information, Innovation, and Competition Policy for the
Internet’ (2013) 161 University of Pennsylvania Law Review 1663; Gabriella Schailer, ‘Electronic
Telecommunications Data as Evidence: Whether the Law has Kept Pace with Data Outputs and
Ordinary Engagement with Technology’ (2016) 41(1) Alternative Law Journal 56.

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT