The Commission should revive and even beef up a proposed law on AI liability that the EU’s executive put on the chopping block earlier this year, urges a Parliament study published today.
The legislative proposal lays down rules on who would have to pay up if an AI system harms someone – and also make AI providers responsible for proving their systems complied with liability rules, instead of putting the burden of proof on people who suffered harm to prove that an AI system had failed.
But all that became moot in February when the Commission announced it would withdraw the proposal, causing consternation among MEPs working on their negotiating position for the file.
Council and big tech lobbies, on the other hand, welcomed the Commission bid to can the law as part of an ongoing regulatory simplification drive. But that’s the wrong way to think about it, per the study commissioned by Parliament and penned by Andrea Bertolini, a law professor at the University of Pisa.
Without an EU law, member countries will just make their own rules – as Germany is a doing for automated driving and Italy is contemplating – the study warns. This would then lead to overregulation by fragmentation, is the concern. “Ultimately, the only way to avoid excessive regulatory burdens is to adopt efficient legal norms,” Bertolini argues.
The study suggests both reviving the AI Liability proposal but also revising it to clarify its scope and rules. It also advocates for opting for a Regulation (rather than the proposed Directive) – so it would be directly applicable in the whole EU – to “deliver the highest degree of harmonization, efficiency, and user protection”.
A prior study, released by Parliament last year, had also recommended the AI Liability law should be a Regulation too.
Sergey Lagodinsky, Green shadow rapporteur on the file, said the latest study confirms the legal affairs committee’s call for a harmonised European framework.
Tiemo Wölken – a substitute member of the committee who is currently suing the Commission over the file’s withdrawal – said the analysis offers a “stark warning” against leaving AI liability unregulated.
“Simply withdrawing the AI Liability Directive, as the Commission intends, is clearly not a viable option,” he also told Euractiv.
The Commission has been contacted for comment
(nl)