Tom Harrington LLB LLM F Inst. MTD, PhD Student, University of Galway
Tom Harrington is a driver training professional with over 50 years of experience in driver education, road safety, and basic, intermediate, advanced and defensive driving. He also trains driving instructors to Road Safety Authority (RSA) standards. In recognition of his contribution to the driver training industry, he was inducted as a Life Member of the prestigious Institute of Master Tutors of Driving, Great Britain.
Tom is currently pursuing a PhD at the University of Galway, with research interests in motoring law and autonomous vehicles. His academic interests include the legal, regulatory and liability challenges arising from emerging vehicle technologies. Combining extensive practical experience with academic research, he brings a valuable industry perspective to the future of motoring law, road safety and autonomous mobility.
Introduction
Autonomous vehicles present fundamental challenges to traditional fault-based liability systems, particularly in Ireland, where no bespoke statutory regime for automated driving has yet been introduced. This article revisits the allocation of liability in autonomous vehicle systems using a game-theoretic framework, situating Irish tort law within an evolving and increasingly integrated European Union (EU) regulatory environment.
In Ireland, liability for road traffic incidents continues to be governed primarily by negligence principles, insurance mechanisms, and the EU-aligned product liability framework.[1] However, significant recent developments at EU level have materially reshaped the regulatory context. The EU Artificial Intelligence Act (Regulation (EU) 2024/1689) classifies autonomous driving systems as high-risk AI, imposing extensive compliance, transparency, and risk management obligations on providers and deployers.[2] These obligations do not directly determine civil liability but operate as regulatory benchmarks that may influence judicial assessments of fault and standard of care.
Further, the revised EU Product Liability Directive (EU 2024/2853) modernises strict liability rules by explicitly extending the definition of “product” to include software and AI systems, including autonomous driving technologies.[3] It also introduces claimant-friendly evidentiary presumptions designed to address informational asymmetries in complex technological claims, thereby lowering the burden of proof in establishing defect and causation.[4] Causation requires proof that the defendant’s breach actually resulted in the damage suffered, usually assessed using the “but for” test. In addition, the law recognises limits where intervening events break the chain of causation or where policy reasons require restricting liability.[5] Damages must also not be too remote, following the principle in The Wagon Mound (No 1), meaning that only reasonably foreseeable types of harm will be recoverable.[6]Taken together, these rules provide a structured yet adaptable framework for determining liability in negligence cases.
Following the European Commission’s withdrawal of the proposed AI Liability Directive in 2025, the EU no longer has a dedicated legislative framework governing fault-based civil liability for AI systems.[7] Instead, liability issues are addressed through the combined operation of the AI Act, the revised Product Liability Directive (EU) 2024/2853, and the existing tort laws of individual Member States. Although these measures promote greater consistency in the regulation of AI across the EU, they fall short of creating a comprehensive and fully harmonised framework for AI-related civil liability.
Against this regulatory backdrop, the article develops a game-theoretic model involving manufacturers, software developers, vehicle operators, insurers, and regulators. It analyses how evolving legal exposure alters strategic behaviour in system design, deployment decisions, and risk allocation. The model demonstrates a structural shift away from driver-centric liability toward a distributed responsibility framework, in which compliance with AI regulation, system transparency, and safety assurance function as strategic signals influencing equilibrium outcomes.
The article concludes that Ireland’s current liability framework, while still grounded in traditional tort and insurance principles, is increasingly shaped by EU regulatory reforms that shift risk exposure upstream toward manufacturers and software developers. Rather than reflecting a legal vacuum, the emerging position is better understood as a layered and adaptive hybrid regime in which liability is progressively redistributed across the autonomous vehicle lifecycle.
1. Background
Autonomous vehicles are transforming not only the technology of driving but also the legal frameworks that underpin responsibility on the roads. In Ireland, liability for road accidents is largely defined by the Road Traffic Act 1961, which assumes that drivers control their vehicles and can be held accountable for their mistakes.[8] AVs, however, shift the locus of control from human drivers to complex software systems, creating new challenges for the application of traditional tort law.
Road traffic accidents remain a significant public safety issue. Statistics indicate that human error contributes to over 90% of accidents in Ireland and across the EU.[9] AVs have the potential to reduce these accidents, yet they also generate complex questions of liability. If an AV causes an accident, is the manufacturer responsible, or does the fault lie with the human occupant? How should liability be distributed when AVs and human-driven vehicles share the same roads? The game-theoretic model developed by Di, Chen, and Talley examines liability as a system of incentives influencing behaviour. It proposes that rational actors—drivers, AV systems, and manufacturers—adjust care based on the potential costs of accidents and liability exposure.[10]
While this framework provides valuable insights into the role of incentives, it does not adequately account for Irish doctrinal law, evidential burdens, or the ethical dimensions of automated decision-making. This article argues that negligence law alone cannot address the complexities of AV liability in Ireland. Negligence continues to occupy a central position within Irish tort law. To establish liability, a claimant must prove that the defendant owed a duty of care, breached the applicable standard of care, caused the alleged harm, and that the claimant suffered actionable damage recognised by law. Although these elements are firmly embedded in Irish law, the courts apply them with considerable flexibility, particularly when determining whether a duty of care exists and whether that duty has been breached.
The modern Irish approach to establishing a duty of care is grounded in policy considerations and was authoritatively articulated by the Supreme Court in Glencar Exploration plc v Mayo County Council (No 2).[11] The Court held that the mere foreseeability of harm is insufficient to give rise to a duty of care. Instead, it must also be shown that there is a sufficiently close relationship between the parties and that imposing such a duty is fair, just, and reasonable in the circumstances. This framework reflects the enduring influence of Donoghue v Stevenson,[12] in which the neighbour principle laid the foundation for the modern law of negligence.
Once a duty of care has been established, the question of breach is determined by reference to the objective standard of the reasonable person. Irish courts have consistently recognised that this standard must be assessed within the factual context of each case. In Ward v McMaster,[13] the Supreme Court confirmed that the defendant’s conduct should be evaluated by considering the circumstances in which it occurred rather than by applying an inflexible or theoretical benchmark. Accordingly, the courts take into account matters such as the degree of foreseeable risk and the precautions that a reasonably prudent person would have taken in similar circumstances.
A hybrid approach is required, combining strict liability for manufacturers with negligence-based liability for human drivers. This model preserves incentives for innovation while ensuring accountability and protecting road users.[14]
2. Game Theory and Liability Incentives
Game theory provides a useful lens for understanding the allocation of liability in AV contexts.[15]According to Di, Chen, and Talley,[16] actors—including drivers, manufacturers, and regulators—make strategic choices about the level of care they exercise, balancing the potential costs of accidents against the likelihood of being held liable. This approach conceptualises liability as a mechanism for shaping behaviour and preventing accidents.
For example, if manufacturers know that they will be strictly liable for AV defects and malfunctions, they are incentivised to implement rigorous safety protocols, conduct comprehensive testing, and ensure that both the vehicle’s hardware and software meet high safety standards. Similarly, human drivers, aware of the legal consequences of inattention or reckless behaviour, may exercise greater care. The model predicts that liability can be optimised through careful alignment of incentives.[17]
However, human behaviour frequently diverges from rational models. Fatigue, distraction, emotional responses, and cognitive bias often result in decisions that are not strictly rational. Irish courts recognise this reality. In McNamara v ESB,[18] the Supreme Court emphasised that legal standards of care must reflect the practicalities of human behaviour rather than idealised expectations. AVs, although precise, are not infallible: software errors and sensor limitations remain unavoidable. Thus, while the game-theoretic framework is analytically useful, it cannot replace legal doctrine or provide a comprehensive solution for allocating liability. It is best used as a tool to complement, not substitute, the legal framework.
3. The Limits of Negligence in Irish Law
Negligence law remains central to Irish tort liability.[19] To establish negligence, claimants must prove duty of care, breach, causation, and damages. The standard is that of a reasonable person.
Irish courts have traditionally interpreted flexibly to account for circumstances. The “reasonable person” standard is an objective measure used in law, especially in negligence, to judge behaviour. It considers what an ordinary, sensible and careful person would have done in the same situation. It ignores the individual characteristics of the defendant and instead applies a common benchmark of expected conduct. Courts use this test to determine whether a person’s actions fell short of the standard of care required, potentially leading to legal liability.
Applying this framework to AVs is challenging. In Glencar Exploration plc v Mayo County Council,[20]the Supreme Court stressed the importance of foreseeability and proximity in establishing duty. But assessing whether an algorithm or AV system has acted “reasonably” presents new difficulties. Courts may lack the technical expertise to evaluate complex machine decision-making, and claimants often do not have access to critical design information.[21]
Further, evidential burdens pose a problem. Human plaintiffs may struggle to demonstrate that a malfunctioning AV caused the accident due to the opacity of “black box” decision-making. Irish law has previously recognised limitations on access to evidence, but AVs exacerbate these challenges. For instance, in Hanrahan v Merck Sharp & Dohme,[22] the court allowed for shifting evidential burdens when defendants had superior knowledge—an approach that may become increasingly relevant in AV cases. Although negligence law has traditionally evolved in response to technological change, its existing principles may be insufficient to resolve the complex questions of responsibility created by autonomous vehicles, where decision-making and operational control are progressively shifting from human drivers[23] to automated systems, manufacturers, and software developers. In the absence of appropriate legal reform, an exclusive reliance on negligence-based liability may generate uncertainty over the allocation of responsibility, hinder effective victim compensation, and create difficulties for manufacturers in assessing their potential exposure to liability.[24] A more comprehensive liability framework, incorporating principles of risk allocation, compensation, and incentives for safer system design,[25] may therefore offer a more effective approach to regulating autonomous vehicle harms.
4. Strict Liability for Manufacturers
Strict liability provides a framework where manufacturers are responsible for defects regardless of fault. EU law already recognises this principle under the Product Liability Directive (85/374/EEC), and Irish courts have applied it in cases such as O’Byrne v Gloucester.[26] The court’s test under the Product Liability Directive, when deciding whether a manufacturer is liable, will generally involve the following questions:
- Was there a product?
The item must fall within the definition of a “product.” This includes manufactured goods, component parts, and, following later amendments, many agricultural products.- Was the product defective?
This is the central test. Article 6 of the Directive provides:
‘A product is defective when it does not provide the safety which a person is entitled to expect’.
- Was the product defective?
In deciding this, the court considers:
- the presentation of the product (warnings, instructions, advertising);
- the use to which it could reasonably be expected to be put; and
- the time when the product was put into circulation.
This is known as the consumer expectations test.
- Did the defect cause the damage?
The claimant must prove that:
- the defect existed; and
- the defect caused the injury or property damage.
Causation remains an essential element, even though negligence does not have to be proved.
- Did the claimant suffer recoverable damage?
Compensation may be awarded for:- death;
- personal injury; and
- certain types of property damage[27]
The EU’s regulatory approach, particularly through the Artificial Intelligence Act, marks a significant shift towards ex ante regulation of high-risk AI systems, including those used in autonomous vehicles.[28] The AI Act adopts a risk-based framework, classifying AV-related systems as “high-risk” due to their potential impact on safety and fundamental rights.[29] This triggers stringent obligations on providers, including requirements relating to data governance, transparency, human oversight, robustness, and cybersecurity.[30]
In practical terms, these obligations are designed to reduce systemic risk before deployment rather than relying solely on ex post liability mechanisms. For example, providers of high-risk AI systems must ensure traceability of decision-making processes, maintain technical documentation, and implement risk management systems throughout the lifecycle of the system.[31] This is particularly relevant in the AV context, where post-accident reconstruction is often hindered by algorithmic opacity.
The AI Act therefore complements, rather than replaces, traditional tort principles. While Irish negligence law focuses on fault and breach after harm has occurred, the AI Act embeds preventative compliance duties at the design and development stage.[32] This dual structure reflects an emerging EU model in which ex ante regulatory compliance requirements indirectly inform the standard of care in subsequent civil liability assessments.
In this sense, the AI Act may influence Irish courts when assessing what constitutes reasonable behaviour under the Donoghue v Stevenson[33] negligence framework, particularly in complex technological environments where the content of the duty of care is informed by regulatory standards. More broadly, EU regulatory initiatives on artificial intelligence emphasise core principles such as transparency, accountability, and structured risk management for high-risk systems, including autonomous vehicles.[34] These measures reinforce the development of a safety-oriented regulatory environment that supports both effective technological innovation and the strengthening of legal protections within national liability frameworks such as Ireland’s tort system.[35]
5. Mixed Traffic and Human Drivers
The coexistence of AVs and human-driven vehicles generates complex liability issues. Humans are inherently inconsistent, whereas AVs operate according to programmed logic. This disparity creates challenges for fair liability allocation. Irish law allows contributory negligence.[36] The O’Sullivan v Dwyer[37] case on contributory negligence and the apportionment of fault. The facts of the case were as follows: the plaintiff was a carpenter who sustained serious injuries after falling while carrying out roofing work for the defendant. He brought proceedings in negligence and for breach of statutory duty.
The jury awarded damages but found that the defendant was 60% at fault and the plaintiff was 40% at fault. The Supreme Court upheld the finding on liability and the apportionment of fault, but considered the award of general damages excessive and ordered a retrial on that issue.
It was heard at first instance in the High Court in Cork. The appeal was determined by the Supreme Court which delivered judgment in 1971. The courts apportioned liability according to relative fault. Yet comparing human error to system malfunction raises fairness and conceptual issues. Should human drivers be held to a higher standard when sharing roads with AVs? Conversely, should manufacturers bear disproportionate responsibility when AVs operate as intended? These questions underscore the limitations of existing frameworks and highlight the need for hybrid liability approaches that address the realities of mixed traffic.[38]
6. Evidence, Information Asymmetry, and Transparency
One of the central challenges in AV liability is information asymmetry. Manufacturers and developers possess detailed knowledge of system design and operation, which is often inaccessible to claimants or courts. Irish courts have occasionally adjusted evidential burdens to reflect such asymmetry. In Hanrahan v Merck Sharp & Dohme,[39] the Supreme Court recognised that the party with superior knowledge may bear a greater evidential responsibility. For AVs, similar principles could require disclosure of software logs, sensor data, and system updates to enable proper adjudication. The Supreme Court rejected the proposition that a defendant’s superior knowledge or technical expertise is sufficient, of itself, to shift the burden of proof. Although the defendant possessed exclusive knowledge of its manufacturing processes, chemical emissions, and pollution control systems, the critical issue of causation remained a matter for scientific and expert evaluation rather than one exclusively within the defendant’s knowledge. The Court declined to formulate a definitive test for superior knowledge, instead affirming that any shift in the evidential burden arises only where the material facts are peculiarly or exclusively within the defendant’s knowledge and cannot reasonably be established by the plaintiff. Consequently, the plaintiffs retained the legal burden of proving causation. The EU AI Act further imposes obligations on high-risk AI systems to ensure risk management, transparency, human oversight, accuracy, robustness and cybersecurity.[40] These regulatory measures complement hybrid liability proposals by providing mechanisms to assess system behaviour and assign responsibility accurately.[41]
7. Insurance and Risk Allocation
Insurance is a key component of road traffic liability. Current systems focus on driver-based insurance, but AVs may require a shift toward product-based coverage. Insurers can influence behaviour through risk-based premiums, reinforcing legal incentives for safety. The EU Motor Insurance Directive guarantees that victims of road traffic accidents receive compensation through compulsory insurance systems, but it is still largely structured around a traditional driver-centred model.[42] This reflects a legal framework designed for human-controlled vehicles, where liability is typically linked to driver fault. However, this approach becomes increasingly difficult to apply in the context of autonomous vehicles, where control is partially or fully transferred to automated systems and software-driven decision-making replaces human input.
To address this mismatch, there is a growing case for developing insurance models tailored specifically to autonomous vehicles.[43] Such models would move away from a focus on individual driver responsibility and instead adopt a broader allocation of risk that reflects the role of manufacturers and system designers. In this regard, combining compulsory insurance with strict liability imposed on manufacturers could provide a more coherent framework for regulating harm caused by autonomous systems. The 1985 Product Liability Directive[44] is relevant because it established the principle that:
- manufacturers are strictly liable for damage caused by defective products;
- an injured person need not prove negligence, only that the product was defective, damage occurred, and there was a causal link between the defect and the damage.[45] Applied to autonomous vehicles, the Directive supports the proposition that where an accident results from a defect in the vehicle’s hardware, software, or AI system, liability may rest with the manufacturer rather than the human driver. However, this raises an important question concerning the level of safety that consumers are entitled to expect from increasingly autonomous vehicles. The concept of defectiveness is particularly significant where an autonomous vehicle is designed to make decisions independently of human intervention. A claimant may reasonably expect a vehicle marketed as autonomous to operate safely without continuous human supervision, but the precise boundaries of that expectation are less clear where autonomous systems are capable of learning, adapting, or producing behaviour that was not specifically anticipated by the manufacturer. The question therefore arises whether the legally expected level of safety should reflect the ordinary expectations applicable to conventional vehicles, or whether greater expectations should arise from the vehicle’s autonomous capabilities and the representations made by its manufacturer.[46]This combined approach would help align incentives more effectively. If manufacturers are aware that they may be held strictly liable for defects or malfunctions in autonomous driving systems, they are likely to prioritise higher safety standards, including more rigorous testing, improved software reliability, stronger cybersecurity measures, and ongoing system monitoring. At the same time, it would enhance protection for victims by ensuring faster and more accessible compensation without the need to establish fault through complex litigation processes. Importantly, such a framework would not necessarily hinder innovation; rather, by distributing risk more efficiently and providing clearer legal certainty, it could support the responsible development and wider adoption of autonomous vehicle technology.
Comparative legal developments further demonstrate this gap. For instance, the United Kingdom’s Automated and Electric Vehicles Act 2018 places primary responsibility for compensation on insurers in cases involving automated vehicles,[47] while allowing insurers to pursue recovery from manufacturers where defects are identified. Similarly, reforms in Germany’s road traffic legislation have adjusted liability rules to accommodate higher levels of automation, distributing responsibility between users and system developers.[48] These examples reflect an international trend towards reallocating liability away from individual drivers and towards product- and system-oriented frameworks.[49]
8. A Hybrid Liability Framework
A hybrid approach combines strict liability for manufacturers with negligence-based liability for human drivers.[50] Manufacturers remain accountable for system failures, while drivers retain responsibility for personal errors. This framework aligns liability with control, enhances safety, and addresses ethical concerns. AVs should not be judged solely by human standards, but accountability must remain meaningful to preserve public confidence in the legal system.[51]
9. Conclusion
Autonomous vehicles expose the structural limits of traditional negligence law. Game theory can explain incentives, but Irish law requires a practical and doctrinally sound framework for allocating liability. Negligence alone cannot address the complexity of AV systems. A hybrid framework, combining strict liability for manufacturers with negligence for human drivers, offers a balanced, innovation-friendly, and accountable solution. Without reform, AVs highlight the limitations inherent in traditional negligence-based liability frameworks.
The EU Motor Insurance Directive ensures that victims of road traffic accidents are compensated through compulsory insurance schemes; however, it remains grounded in a driver-centred structure.[52]This reflects a legal system originally designed for human-operated vehicles, where liability is primarily tied to driver fault. In the context of autonomous vehicles, this model becomes less effective, as control shifts from human drivers to automated systems governed by software-based decision-making.
Against this background, there is increasing support for insurance models specifically designed for autonomous vehicles.[53] Such models would reduce reliance on individual driver fault and instead reflect the role of manufacturers and system designers in generating risk. In this context, combining compulsory insurance with strict manufacturer liability could create a more effective structure for addressing harm caused by autonomous systems.
This approach would also strengthen incentives for safety. If manufacturers face strict liability for defects or malfunctions in autonomous driving systems, they are more likely to invest in rigorous testing, improved software reliability, enhanced cybersecurity, and continuous system monitoring. At the same time, it would improve access to compensation for victims by reducing the need to prove fault through complex litigation. Importantly, such a framework would not necessarily hinder innovation; instead, by clarifying liability and distributing risk more efficiently, it could facilitate the safe development and broader adoption of autonomous vehicle technology.
[1] See Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 [2024] OJ L 1689, arts 6, Annex III, arts 8–15.
[2] ibid Article 6 and Annex III.
[3] Directive (EU) 2024/2853, arts 4(1), 8 and 11; Recitals 13, 32, 50–51.
[4] ibid art 10(2)–(4).
[5] Cunningham v McGrath Bros Ltd [1964] IR 209, 214–15; Breslin v Corcoran [2003] 2 IR 203, 216; Hayes v Minister for Finance [2007] 3 IR 190.
[6] The Wagon Mound (No 1) (Overseas Tankship (UK) Ltd v Morts Dock and Engineering Co Ltd [1961] AC 388 (PC).
[7] European Commission, Proposal for a Directive of the European Parliament and of the Council on Adapting Non-Contractual Civil Liability Rules to Artificial Intelligence (AI Liability Directive) COM(2022) 496 final (28 September 2022).
[8] Road Traffic Act 1961 (Ireland), ss 52–53; Breslin v Corcoran [2003] 2 IR 203, 210–11.
[9] European Commission, ‘EU Road Safety Statistics’ (2021) <https://ec.europa.eu/transport/road_safety> accessed 3 February 2026.
[10] Xuan Di, Xu Chen and Eric L Talley, ‘Liability Design for Autonomous Vehicles and Human-Driven Vehicles: A Hierarchical Game-Theoretic Approach’ (2020) 118 Transportation Research Part C: Emerging Technologies 102710, 27.
[11] [2002] 1 IR 84.
[12] [1932] AC 562.
[13] [1988] IR 337.
[14] European Commission, ‘On the road to automated mobility’ COM (2018) 283 final.
[15] Roger Myerson, Game Theory: Analysis of Conflict (Harvard University Press 1991).
[16] Di, Chen and Talley (n 10).
[17] Richard A Posner, Economic Analysis of Law, (9th edn, Wolters Kluwer 2014) p.185 – 194.
[18] [1975] IR 1 (SC).
[19] Donoghue v Stevenson [1932] AC 562 (HL):, Breslin v Corcoran and Motor Insurers Bureau of Ireland [2003] 2 IR 203, 208–09;, Kelly v Hennessy [1995] 3 IR 253, 257–58.
[20] [2002] 1 IR 84 (SC).
[21] Law Commission of England and Wales & Scottish Law Commission, Automated Vehicles: Summary of Consultation Paper 1 (Law Com No 353, 2022).
[22] [1988] ILRM 629 (SC).
[23] Gary E Marchant and Rachel A Lindor, ‘The Coming Collision Between Autonomous Vehicles and the Liability System’ (2012) 52 Santa Clara Law Review 1321.
[24] Di, Chen and Talley (n 10).
[25] ibid.
[26] [1988] WJSC-SC 2744; see Liability for Defective Products Act 1991 (implementing Directive 85/374/EEC) and relevant Case C-127/04 O’Byrne v Sanofi Pasteur MSD Ltd ECLI:EU:C:2006:93 jurisprudence.
[27]Council Directive 85/374/EEC, art 9(b); Liability for Defective Products Act 1991, s 1(1), s 3. Prop…”
[28] Regulation (EU) 2024/1689 of the European Parliament and of the Council Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act).
[29] ibid, art 6 and Annex III.
[30] ibid.
[31] ibid, art 12.
[32] European Commission, REPORT FROM THE COMMISSION TO THE EUROPEAN PARLIAMENT, THE COUNCIL AND THE EUROPEAN ECONOMIC AND SOCIAL COMMITTEE Report on the safety and liability implications of Artificial Intelligence, the Internet of Things and robotics (2020) COM(2020) 64 final.
[33] [1932] AC 562 (HL).
[34] European Commission, Proposal for a Regulation of the European Parliament and of the Council Laying Down Harmonised Rules on Artificial Intelligence (Artificial Intelligence Act) COM(2021) 206 final, recitals 13–14, 43; arts 9, 13–15.
[35] European Commission, On the road to automated mobility: An EU strategy for mobility of the future COM(2018) 283 final.
[36] Civil Liability Act 1961 (Ireland), s 34.
[37] [1971] IR 275 (SC).
[38] Di, Chen and Talley (n 10); Mark A Geistfeld, Product Liability Law in the Age of AI (3rd edn, Aspen Publishing 2025).
[39] [1988] ILRM 629 (SC).
[40] Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), arts 9, 13–15.
[41] Di, Chen and Talley (n 10).
[42] Directive 2009/103/EC relating to insurance against civil liability in respect of the use of motor vehicles and enforcement of the obligation to insure against such liability.
[43] European Commission, Liability for Artificial Intelligence (2020) COM(2020) 64 final, is the 2020 White Paper/Policy Report on AI liability. It was a Communication that explored options for adapting liability rules to AI. It has not been withdrawn, but it has largely been superseded by subsequent legislative developments, including the proposed AI Liability Directive (2022), the AI Act, and the revised Product Liability Directive.
[44] Council Directive 85/374/EEC on liability for defective products [1985] OJ L210/29.
[45]Council Directive 85/374/EEC of 25 July 1985 on the approximation of the laws, regulations and administrative provisions of the Member States concerning liability for defective products [1985] OJ L210/29, arts 1, 4 and 6.
[46] ibid art 7(1)–(2), especially art 7(2)(b)–(d); see also Recitals 31–32. The Directive provides that a product is defective where it does not provide the safety that a person is entitled to expect, taking into account all circumstances, including the presentation of the product, its reasonably foreseeable use and the time when it was placed on the market or put into service. The Directive also expressly recognises products capable of learning or acquiring new features after being placed on the market.
[47] Automated and Electric Vehicles Act 2018, ss 2–5.
[48] Act Amending the Road Traffic Act and the Compulsory Insurance Act (Germany) (2021).
[49] Bryant Walker Smith, ‘Automated Driving and Product Liability’ (2017) 87 Mississippi Law Journal 277.
[50] European Commission, Proposal for a Directive of the European Parliament and of the Council on Adapting Non-Contractual Civil Liability Rules to Artificial Intelligence (AI Liability Directive) COM(2022) 496 final (28 September 2022).
[51] Di, Chen and Talley (n 10).
[52] Directive 2009/103/EC relating to insurance against civil liability in respect of the use of motor vehicles and enforcement of the obligation to insure against such liability.
[53] European Commission, Liability for Artificial Intelligence (2020) COM(2020) 64 final.
