On Wednesday 18 March, the European Parliament’s Committees on Internal Market (IMCO) and Civil Liberties (LIBE) overwhelmingly approved the proposal for a regulation amending the part of the ‘omnibus’ simplification package dedicated to artificial intelligence.
Adopted by 101 votes in favour, 9 against and 8 abstentions, the draft report introduces strategic deferrals for the application of rules on high-risk systems and seals a major ban on “nudification” tools by AI (see EUROPE 13827/8).
Adjusted timetable. In response to the concerns of the private sector, the MEPs approved the postponement of the activation of the rules relating to high-risk AI systems. The text sets 2 December 2027 as the deadline for systems affecting critical infrastructure, education or law enforcement, a measure that has been widely contested by the European Data Protection Supervisor and the European Data Protection Board (see EUROPE 13792/20).
For products already governed by sector-specific EU legislation, such as medical devices, the deadline has been extended to 2 August 2028.
On the issue of watermarking, however, Parliament has set a faster pace than the Commission, with a deadline of 2 November.
Offensive against sexual deepfakes. One of the most significant additions is a ban on ‘nudification’ systems, which can manipulate images to represent real people in sexually explicit contexts, without their consent. The IMCO and LIBE committees have thus aligned themselves with the Council of the EU, which had included this measure at the request of France and Spain (see EUROPE 13825/5).
Co-rapporteur Michael McNamara (Renew Europe, Irish) strongly supported the measure, pointing out that “being ‘nudified’ and having one’s image abused should not be the price of [...] being a public figure in Europe”. Exceptions to this ban will only be made for systems that prove the effectiveness of their internal security barriers.
Simplification and competitiveness. To lighten the administrative burden of compliance standards for businesses, the text also extends the support measures available to small and medium-sized enterprises to “small mid-cap companies” (with fewer than 750 employees).
MEPs also authorised the processing of sensitive data to correct algorithmic biases, provided that this remains “strictly necessary”.
For co-rapporteur Arba Kokalari (EPP, Swedish), this vote is a strong signal: “This is maybe a small step for Europe to become an ‘AI continent’, but it's a big step to cut bureaucracy for European companies”, she approved.
Despite this broad consensus, the European Conservatives and Reformists Group (ECR) expressed reservations. While supporting the text, the Polish shadow rapporteur Piotr Müller lamented, in a press release published the same day, the rejection of their amendment strengthening the transparency of exchanges between the Commission and the digital giants, believing that “the current compromise does not yet go far enough to strengthen our competitiveness”.
The text must now be validated in plenary on 26 March, before inter-institutional negotiations can begin with the EU Council. (Original version in French by Justine Manaud)