The European Telecommunications Ministers discussed, during an informal remote meeting on 14 October, the proposal for an Artificial Intelligence (AI) Act presented by the European Commission last 21 April (see EUROPE 12811/10).
“We discussed two issues with all the ministers. On the one hand, the need for a unified and systematic approach in the EU to AI and, on the other hand, the implementation of the legislation”, explained the Commissioner for the Internal Market, Thierry Breton.
A strong majority of Member States recognised the need for a risk-based approach, harmonisation of AI rules and the balance between trustworthy technology and support for innovation.
Nevertheless, the Slovenian Presidency of the EU Council will have to work on a number of issues if it is to achieve its goal of presenting a compromise text by December.
Some Member States, such as Estonia and Belgium, considered that the proposed definitions - especially regarding use - were “too broad” and left “too much room for interpretation by the States”. France and Poland also called for ‘high risk’ uses to be clarified.
On the same subject, Lithuania considered that the use of AI for biometric identification requires “further analysis”, arguing for the introduction of a “toolbox” to clarify the conditions for the use of AI in this field.
A rapid evolution
In order to cope with developments in the sector, several Member States indicated the option of using delegated acts. “This is a segment that is accelerating at a speed that we are not used to”, said Italy, summing up the view of some Member States.
In contrast, other countries have been more cautious about this idea, with Estonia and Poland preferring a “single legislative process”.
The issue of small and medium-sized enterprises (SMEs) was also raised on several occasions. On this point, the watchword of a majority of States is simple: avoid an overload that could harm the competitiveness of SMEs.
In this sense, Ireland recalled its commitment to the country of origin principle, which would allow AI service providers to transmit information only to the authorities of its country of establishment. The latter would then be responsible for disseminating them to the competent authorities of the other Member States.
“This avoids working with 27 states, which would create an additional workload and undermine innovation”, summarised Ireland, the headquarter for many major technology companies.
Finally, in addition to the use of ‘regulatory sandboxes’ advocated by some and described by Thierry Breton as “necessary for the EU to become a leader”, insisted Germany as well on the need to pay particular attention to mechanisms already existing at national and European level. (Original version in French by Thomas Mangin)