Greetings from Brussels.

In September 2022, the European Commission published a proposal for an updated Product Liability Directive and a proposal for an AI Liability Directive. The two proposals complement each other and underscore the commission’s strong desire to address the distrust that many in the European Union appear to have about artificial intelligence products. They should also be read in conjunction with the commission’s proposal for an AI Regulation, which sets out a comprehensive framework with strict standards that high-risk AI products need to meet before they can be put on the market in the EU.

In the latest set of proposals, the commission wants to make clear that consumers and other users of AI products who want to bring a claim for damages caused by these products can do so under the favorable, plaintiff-friendly rules of evidence contained in the current Product Liability Directive.

The current Product Liability Directive

The current Product Liability Directive has been in existence since 1985. It is widely seen as an important instrument for individuals who try to obtain compensation from the manufacturer for damage caused by a defective product. The directive foresees a no-fault regime: the claimant must “only” prove that the purchased product is defective, that it caused damages, and that there is a causal link between the defect and the damages. The directive substantially eases the burden of proof for claimants by providing a rebuttable presumption that a product is defective if certain conditions are met, including when the claimant establishes that the product does not comply with mandatory safety requirements or that the damage was caused by an obvious malfunction of the product during normal use or under ordinary circumstances. It also provides claimants with a right of access to relevant evidence from the defendant. Failure to provide this evidence by the defendant, when ordered by a court, triggers the presumption that a product is defective.

The revised Product Liability Directive

Although the revised Product Liability Directive will not change the basic approach of the existing directive, it will bring much needed updates to it: a widening of its scope and a significant increase in the number of operators that can be held liable for defective products. Importantly, the revised directive now clearly states that software and AI systems are within scope and that compensation under the directive and its no-fault approach is possible when products like robots, drones or smart-home systems are made unsafe by defective software updates. In other words, if an AI system is defective and causes death, personal injury, property damage or data loss, injured people can use the revised Product Liability Directive to claim compensation.

Why an additional AI Liability Directive?

Notwithstanding the expanded scope of the revised Product Liability Directive, not every type of damage caused by an AI system is covered. Some damages, such as those caused by discrimination or violation of fundamental rights, are out of the directive's scope. The commission provides the example of an applicant who estimates that he was wrongfully denied a position because of a bias in the AI system used during the application process. That applicant cannot file a claim for compensation under the Product Liability Directive and would thus need to resort to the more traditional fault-based, noncontractual liability principles, which can be found in the civil codes of the various EU member states. 

They are not harmonized at the EU level, but they tend to require claimants to prove fault, damage and a causal link between the two. This is never an easy task but is even more challenging when AI systems are involved. The claimant may bear substantial upfront costs in trying to gather the necessary evidence.

This is where the proposed AI Liability Directive comes in. The Directive would introduce for AI-related claims the same alleviation to the burden of proof that the Product Liability Directive does. It uses the same tools — right to disclosure of evidence and rebuttable presumptions — and similar wording to ensure consistency between the two directives. With the introduction of the AI Liability Directive, the differences between AI-related claims bought under the strict liability regime of the Product Liability Directive and claims brought under the more traditional fault-based liability regimes are significantly reduced.

Next steps

Both proposed directives have been sent to the European Parliament and the Council for review and further discussion. Once approved, they will need to be transposed by the member states into their national laws. Member states would have 12 months to do so for the revised Product Liability Directive and 24 months for the AI Liability Directive. Those timelines may still change during the further deliberation of the proposals in the Parliament and the Council. The lengthy EU legislative process and the additional transition time mean that it will still be some time before these two proposals are in force in the various EU member states.

It will be interesting to see how the proposals are received in the European Parliament and the Council. There can be little doubt that the success of the AI-based economy depends at least to some extent on the acceptance of this type of tech by consumers and other users. Having a robust regulatory framework to screen products before they enter the market is essential in this respect, and so is having an efficient redress mechanism for consumers and others in case damages nevertheless occur. That is precisely what the two proposals aim to do.