Proposed liability rules aim to shape AI responsibilities

Debate will resume in the European Parliament this month about who should be held accountable in cases when AI systems cause harm or damage. MEPs are set to discuss the proposal adapting non-contractual civil liability rules to AI (AI Liability Directive or AILD), first tabled by the Commission in September 2022 but put on hold during the negotiations on the Artificial Intelligence Act which was finally published in the EU Official Journal in July 2024 and applies gradually from February 2025 onwards.

The proposed AI liability rules aim to address the risks generated by specific uses of AI through a set of rules focusing on respect for fundamental rights and safety.

According to the European Commission, the proposed directive will introduce uniform rules to ensure that victims of AI-related harm receive the same level of protection as those harmed by other technologies. The proposal eases the burden of proof for victims by allowing a rebuttable presumption of causality, meaning if a victim can show that an AI system likely caused the damage, the defendant must prove otherwise. Additionally, it allows national courts to order the disclosure of evidence about high-risk AI systems to help victims prove their cases.

Background and initial reception

The AILD proposal is intended to complement the recently finalised AI Act, which has delayed substantial work on the directive by both the European Parliament and EU governments in the Council. Since the presentation of the AILD proposal, there has been a general lack of interest from EU governments, with diplomats from several countries expressing indifference towards the file. Discussions have emerged about whether it would make sense to have rules on AI liability, given that EU legislators have simultaneously revised the EU Product Liability Directive to include software.

In early July 2024, Axel Voss, a center-right German lawmaker leading the work on AILD within the Parliament’s Legal Affairs Committee (JURI), expressed concerns about potential overlaps with the new EU product liability rules. Nonetheless, Voss indicated confidence that there would be enough "loopholes" to justify proceeding with the AILD.

Key updates and provisions

Despite initial resistance, the European Commission has remained committed to advancing the AILD. At the end of July 2024, the Commission sent an updated version of its proposal to both the European Parliament and the Council, aligning the wording with the finalised AI Act and introducing tweaks to increase the responsibility of companies deploying AI systems.

The updated AILD proposal includes significant changes, particularly in Article 4, which now presumes liability for companies if they fail to monitor the operation of AI systems or use insufficiently representative input data. This shift aims to ensure that companies deploying AI systems are held accountable for any damage caused by their products.

The European Commission's dedication to the AILD was reiterated during an exchange in the Parliament's legal affairs committee in December 2024, where a representative expressed the executive's eagerness to see discussions move forward swiftly. MEP Voss also emphasised the need to continue working on the directive, highlighting gaps that still need to be addressed.

Parliamentary and council dynamics

Before the summer of 2024, the Legal Affairs Committee, responsible for the file in Parliament, requested the European Parliamentary Research Service (EPRS) to conduct a complementary impact assessment of the proposal. The study, published in September 2024, identifies key shortcomings in the European Commission's impact assessment, particularly regarding the strict liability regime. The complementary impact assessment study proposes that the AILD should extend its scope to include general-purpose and other 'high-impact AI systems', as well as software. It also discusses a mixed liability framework that balances fault-based and strict liability.

While the JURI Committee seems intent on proceeding with the examination of the file, resistance within the Parliament is growing, with some members questioning the necessity of the directive in light of the EU's renewed competitiveness drive. Christian-Democrat MEPs Arba Kokolari of Sweden and Andreas Schwab of Germany have raised concerns about the potential disproportionate impact on European small- and medium-sized enterprises and startups.

In the Council, discussions have been challenging, with the Hungarian presidency preparing a discussion paper focusing on key articles of the AILD. The paper emphasises the need for clear liability rules to prevent uncertainty from hindering AI adoption.

Moving forward

Despite pushback from several member states, the Parliament appears interested in resuming discussions on the AILD and adopting a position. 

The Internal Market and Consumer Protection Committee (IMCO) has been reappointed as the Committee for Opinion on the file, with a detailed timetable for the adoption of the opinion already in place. MEP Kosma Złotowski (ECR, Poland) will draft the opinion, which should be ready by the end of January. A vote is expected in May 2025.

Within the responsible European Parliament’s Legal Affairs Committee (JURI), MEP Voss aims to publish his draft Report on 4 June 2025. This is expected to be followed by a vote in the JURI Committee in January 2026 and a vote in the Parliament’s plenary session in February 2026. 

The Polish Presidency of the Council has also expressed interest in advancing work on the proposal during its six-month tenure (January – June 2025).

The upcoming months will therefore be crucial for the fate of the AILD, as the Parliament and Council work towards finding common ground on the directive. The European Commission's commitment to the AILD, coupled with the Parliament's intent to deliver on the rules, suggests that progress is likely in 2025, albeit with potential adjustments to address the concerns raised by various stakeholders.