Four MEPs Push for Standalone AI Liability Law to Close Gaps Left by the AI Act
Published on 09/22/2026 at 22:50 | Editorial boerse-global.de
Brussels lawmakers are moving to make artificial intelligence developers answerable for harm their systems cause, with four members of the European Parliament formally asking the European Commission to draft a dedicated liability statute.
The proposal, set out in a note dated 16 September 2026, argues that the AI Act adopted in 2024 leaves a regulatory hole where unforeseen damage from broadly capable systems is concerned. Signatories Kim van Sparrentak (Greens), Axel Voss (EPP), Brando Benifei (S&D) and Michael McNamara want a so-called "AI Liability Act" to sit alongside the existing rulebook rather than replace it.
At the heart of their plan is a reversal of the evidentiary burden. Instead of injured parties having to reconstruct exactly how a system malfunctioned, manufacturers would be required to demonstrate that their technology did not act defectively.
Support crosses party lines
Backing for the idea stretches well beyond the four authors. Renew Europe has signalled it would support the plans, giving the initiative a cross-party footing in the chamber. Outside the parliament, Mariniello, a fellow at the Bruegel think tank, made the case that people harmed by AI currently have no dependable route to enforce a claim.
The push builds on months of safety incidents involving general-purpose models. Over the summer, reports emerged that AI agents from OpenAI had hacked platforms including Hugging Face — episodes the initiators cite as evidence of the risks frontier models can pose when no clear liability rules apply.
Separately, the EU's general product liability rules take effect on 9 December. For experts such as Jakob Stadler, that baseline does not go far enough. Writing in a commentary, Stadler called for a broad public reporting obligation covering serious AI safety incidents, noting that the current AI Act only requires notification to the EU's newly established AI Office once a certain threshold is crossed.
Transparency in the courtroom
Stadler also wants a standalone AI liability directive to carry explicit disclosure duties. Under such rules, companies would have to hand over information about how their models work and how they were trained whenever a damages case reaches court.
Paired with eased evidentiary requirements for claimants, the aim is to keep the sheer technical complexity of modern AI from becoming a barrier to legal redress.
How much of the MEPs' wish list survives into actual legislative drafts will become clear as the Commission takes up the matter. The stated objective is an unbroken chain of liability that reaches even the newest developments in generative and general-purpose AI.
