European Commission
The AI Liability Directive (AILD), proposed by the European Commission in September 2022, is a key element of the European Union's ongoing efforts to reform AI liability regulations. As part of a broader AI policy framework complementing the proposed AI Act, the AILD aims to introduce fault-based civil liability provisions for AI-related damages. The directive's main objective is to ensure that individuals affected by AI systems enjoy the same level of protection as those impacted by other technologies within the EU. To achieve this, the AILD seeks to harmonise non-contractual civil liability rules and simplify consumers' ability to pursue compensation for AI-related product and service-related damages. The proposed directive is currently undergoing the legislative process, and after adoption by the EU institutions, the directive must be implemented in local EU Member States' legislation within two years time. It is important to note that the AI Liability Directive is classified as a Directive. Due to its nature, the Directive sets specific goals that all EU Member States are mandated to achieve. However, the precise application and rules stemming from a directive vary, as each Member State has the discretion to choose how to integrate the directive's goals into its own national law. To fully understand the rights and obligations under a Directive, it is essential to examine the specific national law into which the Directive has been incorporated in the relevant EU Member State.
The AILD applies to non-contractual fault-based civil law claims for damages caused by an AI system in cases where the damage occurs after the end of the transposition period of the Directive. As the AILD concerns non-contractual liability claims, no contractual link between the party suffering the damage and the liable person is needed in order to claim damage. The proposed directive does not provide the definitions for ‘AI system’, ‘provider’, or ‘user’ but rather refers to the definitions laid down in the proposed AI Act. Claim for damages is defined in the directive as non-contractual fault-based civil law claim for compensation of damage. The damage can be either caused by an output of an AI system or alternatively the failure of an AI system to produce an output, where such output should have been produced. This means that victims, whether they are natural or legal persons, can seek compensation for harm suffered due to fault or omission of the defendant where the harm results in damage. It is important to note that the scope of the damage applies to any type of AI system, meaning that all AI systems irrespective of their risk classification under the AI Act, fall within the scope of the proposed AILD.
The AILD lays down standard rules on the disclosure of evidence on high-risk artificial intelligence (AI) systems to enable a claimant to substantiate a non-contractual fault-based civil law claim for damages and on the burden of proof in the case of non-contractual fault-based civil law claims brought before national courts for damages caused by an AI system. The AILD does not apply to criminal liability.
This Directive will not affect: (a) rules of Union law regulating conditions of liability in the field of transport; (b) any rights which an injured person may have under national rules. (c) the exemptions from liability and the due diligence obligations as laid down in [the Digital Services Act] and (d) national rules determining which party has the burden of proof, which degree of certainty is required as regards the standard of proof, or how fault is defined, other than in respect of what is provided for in Articles 3 and 4. Member States may adopt or maintain national rules that are more favourable for claimants to substantiate a non-contractual civil law claim for damages caused by an AI system, provided such rules are compatible with Union law. The European Commission (EC) will review the application of the Directive five years after the end of its transposition period to assess and ensure its effectiveness and suitability.
