German MEP Axel Voss (EPP) spoke to members of the European Parliament’s Committee on Legal Affairs (JURI) on Monday 28 March about legislation on artificial intelligence (AI). These discussions were the second to be held in the ‘JURI’ Committee, which is associated with this subject but has certain exclusive powers.
Several issues should be discussed “intensively” in the JURI Committee in the future, said Mr Voss. The German MEP believes that it will be necessary to focus on definitional issues first. In this respect, “we will not have to extend the concept to traditional systems. A balanced approach is needed”, he said.
In addition, he added, the question of the risks which the text should protect against should also be at the heart of the discussions.
“We need to know what risks we are talking about. Constitutional issues? We need to know what we want to protect against”, he summarised, re-emphasising the need not to include too many AI systems in the ‘high risk’ category.
In addition, the rapporteur for the JURI Committee - in which some 600 amendments have been tabled on AI legislation - also foresees that part of the future exchanges will concern the issue of potential duplication with certain provisions included, among others, in the texts on digital services legislation (DSA) (see EUROPE 12913/20), data legislation (‘Data Act’) (see EUROPE 12913/21) or cybersecurity (see EUROPE 12903/18).
Stressing that an “effective” AI system that avoids “bias and discrimination” cannot be developed without “quality data”, the rapporteur also questioned the link of the future text with the General Data Protection Regulation (GDPR). Mr Voss suggested a number of ways of thinking about this, including the possible use of regulatory ‘sandboxes’.
“The report contains things that go against ‘trustworthy’, with the opening of citizens’ data without their consent”, said Pascal Durand (Renew Europe, France), speaking on behalf of his Danish colleague, Karen Melchior.
The issue of liability of AI system providers was also raised. On this point, the European Commission would like to see this aspect addressed in a proposal “later this year” on the revision of the Product Liability Directive.
This does not reassure Mr Voss, who would prefer to solve the problem as soon as possible. “We have requirements on AI and only then should we talk about liability? This is not very sensible”, he explained, saying that the Commission’s proposal would come “too late”.
Finally, the rapporteur for the JURI Committee also noted that the question of the approach in terms of governance, through the creation of a new agency or the strengthening of existing structures, should also be decided. (Original version in French by Thomas Mangin)