Skip to content

European Commission proposes AI Liability Directive and modernised Product Liability Directive

On 28 September 2022, the European Commission published two proposals aimed at modernising product liability rules in the digital age. The proposal is the first specifically designed to address compensation for damage caused by artificial intelligence (AI) systems. The general product liability rules are also simplified and expanded to address digital products and the circular economy (e.g. repairing, reusing and recycling). 

Two Directives are proposed:

  • One on adapting non-contractual civil liability rules to AI (AI Liability Directive); and
  • The other for defective products (Product Liability Directive).

The proposed Directives are said to complement each other and the EU AI Act (our article on the same can be found here). The Product Liability Directive addresses a producer’s “no fault” or “strict” liability for defective products (including AI) and associated compensation for damages. Whereas, the AI Liability Directive is concerned with liability arising from “wrongful behaviour” by AI systems.

The AI Liability Directive

The AI Liability Directive has two new features which are aimed at tackling some of the challenges faced when dealing with liability stemming from AI:

  • A right of access to evidence. With a traditional product it may be readily apparent what has gone wrong in the event of a defect. With AI this is may well not be the case. To address this, subject to certain conditions, a provider of a high-risk AI system (defined in the EU AI Act) can be ordered by a court to disclose relevant and necessary evidence about their product. In limited cases, requests can also be made from third parties;
  • A rebuttable presumption of causality. Famously when DeepMind’s AlphaGo defeated a human, some of the moves were ones that professional Go players had never considered before. This is an illustration of both the opportunities and challenges posed by AI. As humans, we may not be capable of understanding the logic behind an AI decision or how one thing led to another. This is one of the reasons that legislation relating to AI liability may have to take a policy decision based on a presumption. Accordingly, subject to certain conditions and in narrow circumstances, national courts would be required to presume, for the purposes of applying liability rules to a claim for damages, that the output produced by the AI system (or the failure of the AI system to produce an output) was caused by, for example, the fault of the AI provider.
  • Who can be sued? There has been a lot of debate as to who should be accountable in the event of a failure by an AI system. These proposals conclude it should be the providers of AI systems and, in some cases, the user of AI systems (each as defined in the EU AI Act).

The Product Liability Directive

The revised Product Liability Directive proposal adapts existing product liability rules to address new types of products and services, such as software or AI systems or advanced machinery. Key features include:

  • a right to bring a claim for damages against the manufacturer of a defective product that caused death, personal injury (including medically recognised psychological harm), damage to property or data loss or corruption;
  • software or AI systems and AI-enabled goods are explicitly named as “products” under this proposal, and injured natural persons can claim compensation if software or AI cause damage;
  • in cases where the court considers that a person bringing a claim faces difficulty in proving a defect or causality between defect and damage due to technical or scientific complexity, the burden of proof will be reduced (though the person sued can contest this). This proposal to shift the burden of proof aligns the proposed Product Liability Directive with the AI Liability Directive;
  • not only hardware manufacturers but also software providers and providers of digital services that affect how a product works (by way of example, the provider of a navigation service in an autonomous vehicle) can be held liable;
  • manufacturers can be held liable for changes they make to products they have already placed on the market (eg when these changes are triggered by software updates or machine learning). A person modifying a product that has been placed on the market or put into service may, in some cases, be considered manufacturer and therefore liable for damage;
  • claim thresholds and caps on compensation levels are removed. Under current rules, to come into consideration for a claim, the damage caused by the defective product must be at least EUR 500, and Member States may impose a limit on the producer's total liability for damage resulting from death or personal injury (this amount cannot be lower than EUR 70 million);
  • other parties such as importers, authorised representatives, fulfilment service providers or distributors can be held liable for defective products manufactured outside the EU or where a manufacturer cannot be identified.

The press release is available here. The AI Liability Directive is available here and the Q&A on it, here. The Product Liability Directive is available here and the Q&A on it, here