MEPs on the committees on Civil Liberties, Justice and Home Affairs (LIBE) and Internal Market and Consumer Protection (IMCO) discussed and debated on Wednesday 11 May the report by Dragoș Tudorache (Renew Europe, Romania) and Brando Benifei (S&D, Italy) on the Artificial Intelligence (AI Act) legislation (see EUROPE 12947/2).
The report presented on Wednesday introduces several changes to the European Commission’s original proposal. While the text retains the human-based approach, the Parliament relies on the broadest possible definition of AI systems in terms of scope.
“A narrow definition would leave gaps and AI could be used without regulatory oversight”, Tudorache explained.
The text focuses largely on prohibited systems, those that should be required to be transparent about confidentiality and those that are high risk or ‘dangerous’.
The list of high-risk and dangerous systems has been expanded to include, for example, AI systems used by political parties or personalities to influence people’s voting, systems intended for children, or ‘AI Office’ systems, which are capable of writing informative or scientific articles.
A future fundamental rights impact assessment
In addition, Mr Tudorache called on his colleagues to support amendments to ban AI systems that enable predictive checks. Several members of the S&D group have already announced that they will present amendments to ban biometric recognition in public spaces.
“We have taken note of some of the arguments, particularly from the EU Council, which were aimed at excluding certain types of AI from regulation. We do not agree. AI vendors who produce systems for general purposes need not worry and will not have to bear any ‘regulatory burden’”, insisted Tudorache.
In addition, the report also foresees an increase in the liability of AI system providers and a simplification of the redress process for users in case of problems and notifications from providers.
The issue of respect for fundamental rights was also raised several times. On this point, Brando Benifei announced that he would ask for an impact study to better take into account vulnerable people and groups.
The text also calls for better cooperation between the various stakeholders and wishes to strengthen the role of the future European Council for AI and integrate the EU Fundamental Rights Agency.
As for the Parliament, while the work done was mostly welcomed, several MEPs called for caution to ensure that regulation does not end up weighing down innovation.
“We must weigh the pros and cons so as not to hinder the development of AI. We should focus more on autonomous systems”, said Axel Voss (EPP, Germany).
For others, like Deirdre Clune (EPP, Ireland), some definitions should also be reworked. “There have been reactions from industry, from academics, who say that the interpretation is complicated. This should be reviewed”, she said. (Original version in French by Thomas Mangin)