Series
Blogs
Series
Blogs
Nearly two years have passed since the EU Commission submitted its proposal for an EU AI Liability Directive. In light of significant developments in the legislative landscape since then, the EU Parliament and the Council question whether this legislative project remains suitable. Given the Parliament’s and the Council’s scepticism, the Commission's forthcoming strategy in this area remains uncertain.
Back in September 2022, the EU Commission tabled a proposal for an AI Liability Directive. Initially, with the submission of this proposal, the Commission sought to ensure that individuals affected by artificial intelligence systems receive an equivalent level of protection to those impacted by other technologies. In an effort to make it easier for individuals to bring claims for harm caused by AI, the proposal suggested to introduce measures to alleviate issues of evidence, such as specific rules on disclosure of evidence and a new rebuttable presumption of causality (for more information, see our previous blog post).
It is important to recognize that this proposal did not arrive as an isolated piece of legislation. Rather, it formed part of the EU’s broader effort to update and adapt the product liability regime to meet the challenges posed by the rapidly advancing AI landscape and to foster trust in AI systems. The proposal aimed at a targeted reform of national fault-based liability regimes (which exist as a “second pillar” next to the harmonised EU product liability rules) to ensure that victims of AI related damages benefit from a presumption that applies to all types of claims related to malfunctioning AI systems. The proposal was introduced concurrently with the Commission’s proposals for the AI Act and the Product Liability Directive. While the legislative processes for the latter two initiatives are progressing successfully – the AI Act was officially adopted on 21 May 2024 (read more in our blog post) and the Product Liability Directive is anticipated to be formally adopted by the Council shortly (read more in our blog post) – the future of the AI Liability Directive remains uncertain. In March 2024, the European Parliament adopted, at first reading, its position on the AI Liability Directive, but since then the legislative train stopped.
Given the fact that the AI Liability Directive was intended to apply alongside the AI Act and the Product Liability Directive, its effectiveness depends on the broader legislative framework. This includes, in particular, the legal definition of “AI systems” as well as the classification of certain systems as “high-risk” provided for in the AI Act. These key concepts were subject to significant amendments and negotiation during the legislative process, which must be accurately reflected in the AI Liability Directive as well.
Moreover, the revised Product Liability Directive now explicitly encompasses software and software developers and addresses damages resulting from data destruction or corruption. This update raises concerns about whether the AI Liability Directive offers a route for redress that, in fact, overlaps with the one provided by the Product Liability Directive.
In response to these developments, the EU Parliament has called for a thorough impact assessment of the draft proposal to evaluate its relevance and effectiveness in the current legislative landscape. This assessment, expected to be completed by the end of September, will determine if there are any legislative gaps that the AI Liability Directive could still address following the adoption of the AI Act and the Product Liability Directive. Should the assessment be unfavourable, the Parliament is likely to recommend that the Commission withdraws the proposal.
Similarly, the Council’s enthusiasm for advancing this Directive appears limited. The upcoming Hungarian Presidency has indicated that a general approach from the Council on this act is unlikely in the near term. Several Member States have voiced doubts about the Directive's value, noting that substantial additional work is required to align the Directive's content with the evolving broader legislative framework.
The future of the AI Liability Directive is uncertain. In light of the criticism, the Commission may very well be inclined to withdraw the existing, outpaced proposal and draft a new one that aligns better with the current legislative environment. However, with a view to the imminent EU elections, any decision on this matter is unlikely to be taken before early 2025.