Negotiators from the EU Council and the European Parliament will be attempting to make progress at a new technical meeting on Tuesday 14 November on legislation on artificial intelligence (‘AI Act’).
General-purpose AI systems and foundation models will be at the centre of discussions as several Member States, including France and Germany, have opposed the tiered approach of the Spanish Presidency of the EU Council, which proposed opting for the creation of stricter criteria and rules for ‘high-impact’ foundation models (see EUROPE 13279/18).
Several Member States, including France and Germany, felt that the legislation should be based on an approach focusing on the level of risk posed by AI systems rather than AI foundation models.
The Spanish Presidency of the Council of the EU also proposed abolishing exemptions for AI models for which the source codes are freely available (‘open source’).
While progress on the dossier is slow, the co-legislators are stepping up their meetings ahead of the next round of inter-institutional negotiations (‘trilogue’), which will take place on 6 December. Technical meetings will be held on 16, 24, 27 and 29 November and on 1 and 4 December (see EUROPE 13283/22). Two other technical meetings are already planned after the next trilogue, on 11 and 13 December.
For the time being, the co-legislators have managed to agree on a number of points, such as obligations for suppliers of high-risk AI systems, notifications to the competent authorities, ‘regulatory sandboxes’ and penalties. (Original version in French by Thomas Mangin)