Liability and artificial intelligence in the EU: Assessing the adequacy of the current Product Liability Directive

Abstract

The European Union selected achieving a leadership position in the AI sector as one of the priorities for the future of the bloc as a whole. Economic reasons are behind this decision, but they are not the exclusive motive behind this objective. Undeniably, AI will have an enormous impact on world’s economy and if the EU falls behind, the standards of living that citizens of the Union currently enjoy may be at risk. Furthermore, there is also the question of keeping European values, principles and ethical standards alive in this technological transition. To achieve the leadership position it desires, it is essential for the EU to possess an updated, producer-friendly legal framework, that manages, at the same time, to ensure consumer protection and safe development of AI. One of the legal instruments that may need to be amended is the Product Liability Directive. In this paper we will study the Product Liability Directive and its shortcomings on AI regulation, along with the possible solutions to adapt the EU product liability legislation to this new technological challenge. We will assess what solutions are best suited to apply in the EU and fulfil the objective of achieving leadership in the sector.

A review of product safety regulations in the European Union

“…For these and other reasons, liability for new technologies has been actively debated recently. Given that reforms are already underway on this front, it suffices to generally pinpoint the many challenges: unclear definitions, including the legal status of artificial intelligence systems [29], different legal notions of liability, potential (compulsory) insurance schemes, impacts upon economy and innovation, supply chains, and so on [8,10,75]. Another related point is the ethical use of artificial intelligence promoted by the EU.…”

“…Many of the established demarcations seem outdated, both in academia and in practice. For instance, the distinction between products and services has acknowledgedly become blurry for legal, business, and safety considerations [3,10,21,23]. Particularly information technology products have long relied on different services, nowadays often fueled by data.…”

“…Although the question has been more about security than safety-a distinction, which, as noted, is not always clear, software liability has not generally progressed despite a long global debate [51,65]. Europe is no exception [10]. As will soon be discussed, software and information technology products and services have also introduced other major security and safety challenges.…”

The European Commission’s Approach To Extra-Contractual Liability and AI – A First Analysis and Evaluation of the Two Proposals

“…Moreover, 'hybrids between products and services will be quite frequent'. 230 As outlined above, the revised PLD answers some of these questions by extending the scope of application rationae materiae.…”

The European Approach to Building AI Policy and Governance: A Haven for Bureaucrats or Innovators?

“…The rapid adoption of the technology in the public sector has led to inevitable mistakes that proved that the risks of predictions provided a black box fed by human input. Apart from the notoriously dystopian results of the AI-backed COMPAS system (Black & Murray, 2019), we have to acknowledge the fact that algorithms are powerful enough to transform any sphere of human activity into a race for higher efficiency, be it anti-plagiarism plugins (Büttner et al, 2015) or weaponry (Cabral, 2020). These and other risks have triggered the discussion on how to elaborate a human-centred approach to developing these technologies, which coined terms like XAI (eXplainable AI) (Arrieta et al, 2020) and responsible AI (CEN-CENELEC, 2020).…”