On Wednesday 5 July, at a meeting of the Working Party on Telecommunications, the EU Member States will take stock of the EU Council’s position on artificial intelligence (AI) (see EUROPE 13201/1). This will enable the Spanish Presidency of the EU Council, whose mandate began on 1 July, to sound out the positions of the Member States ahead of the next round of inter-institutional negotiations (‘trilogue’), which will take place on 18 July (see EUROPE 13204/27).
The Spanish Presidency of the EU Council is expected to return initially to the question of the definition of AI systems. On this point, the European Parliament has aligned its definition with that of the OECD (see EUROPE 13155/5). The EU Council has retained only certain elements of the latter. Several options have been put forward by the Presidency of the EU Council, including moving closer to the European Parliament’s position, maintaining the EU Council’s position, or awaiting future developments and OECD guidelines on the subject.
The question of the obligation to carry out an impact study on fundamental rights before a high-risk AI system is placed on the market, which is dear to the hearts of certain members of the European Parliament, should also be discussed. The EU Council’s approach does not include such an obligation and sticks to the provisions included in the General Data Protection Regulation (GDPR). These require companies and public organisations to examine whether there are high risks to rights and freedoms when personal data is processed.
On this point, the new Spanish Presidency of the EU Council is expected to propose a number of avenues for Member States to explore, including the possibility of moving closer to the European Parliament while at the same time doing away with the six-week consultation period requested by MEPs prior to the launch of an impact study.
The question of the classification of high-risk systems, which had been the subject of heated debate - both in the European Parliament and the EU Council - should also be one of the issues raised. The EU Council’s position focuses on the results of artificial intelligence systems when it comes to decision-making, while the European Parliament would like to oblige AI system providers to assess the risks posed by their systems. If, according to the companies and entities concerned, these risks should not cause the system developed to fall into the high-risk category, it would be up to the competent authority to decide.
Several avenues are being examined by the EU Council. One option would be to adopt the European Parliament’s position, while the other would be to retain the approach defined by the EU Council when it adopted its position (see EUROPE 13078/12). However, the Spanish Presidency of the EU Council could also propose finding common ground, for example through binding rules based on clear risk criteria that companies would have to assess. (Original version in French by Thomas Mangin)