The new impact assessment on the EU's AI Liability Directive proposal: an uncertain future ahead

In September 2022, the European Commission put forward a proposal for a directive to revise non-contractual civil liability rules in relation to artificial intelligence (AILD), alongside an impact assessment.

The directive proposal aims at ensuring that individuals damaged by AI systems are granted the same level of protection as those damaged by other technologies across the EU. Additionally, it seeks to ease the burden of proof for those claiming compensation for AI-related damages by introducing mechanisms such as evidence disclosure for high-risk AI systems and rebuttable presumptions. For more details on this, please refer to our previous article here.

Crucially, the AI Liability Directive is designed to operate in conjunction with the AI Act and the Product Liability Directive. However, without proper coordination within the broader legislative framework, there is a risk of overlap or inconsistencies. One clear example is that the AILD proposal does not take into account certain elements that are central to the AI Act, such as the existence of prohibited AI practices or the need for human oversight on high-risk AI systems.

In response to these developments, the European Parliament requested a complementary impact assessment to reassess the relevance and potential of the AILD proposal in today’s legislative context.

This comprehensive assessment, recently published, leaves no area unexplored. It suggests that the AILD should broaden its coverage to include general-purpose AI systems and "high-impact" AI systems, as well as software. It also introduces the idea of a mixed liability framework, combining fault-based and strict liability. The report goes further to recommend transitioning from an AI-specific directive to a broader software liability regulation, in order to promote legal consistency throughout the EU. In the aftermath of this new assessment, the future of the AILD proposal remains uncertain.

You can read the full impact assessment via the following link.

Authored by Christian Di Mauro and Guido Di Stefano.