The Council of the EU adopted its position (‘general approach’) for the future inter-institutional negotiations on artificial intelligence (AI) legislation at a ‘Telecommunications’ Council on Tuesday 6 December (see EUROPE 13055/2).
“Today we have managed to strike a delicate balance that will stimulate innovation and the adoption of artificial intelligence technology across Europe. With all its advantages, on the one hand, and full respect for the fundamental rights of our citizens, on the other”, said Czech Deputy Prime Minister for Digitisation Ivan Bartoš.
In addition to the definition of AI systems, restricted by the EU Council to systems developed by machine learning and logic and knowledge-based approaches, the text also covers prohibited practices in the field of artificial intelligence.
In this respect, the EU Council’s general approach is that social rating should be prohibited. This provision has been extended to private actors. The provision to prohibit the use of AI systems that exploit the vulnerabilities of a specific group of people has been extended to people who are vulnerable because of their social or economic situation.
The text, which has been at the heart of discussions since the beginning of the work on AI, does not prohibit, but defines, on the other hand, reasons and objectives that could legitimise the ‘exceptional’ use of real-time remote biometric identification systems in public space.
Exclusion of national security, defence and military objectives
Another key issue in the discussions - both in the Parliament and in the EU Council - is that the text also covers high-risk AI systems. The EU Council provides that AI systems that are not “likely to cause serious violations of fundamental rights” or other significant risks will not be affected.
Several requirements for these high-risk systems have been clarified, such as data quality or the technical documentation that SMEs must produce to demonstrate that their high-risk AI systems comply with the requirements.
In addition, the text now clarifies that the requirements for high-risk AI systems also apply to general-purpose AI systems in cases where the latter are integrated into high-risk systems. However, on this point, the Council of the EU prefers that an implementing act specifies how they are to be applied, based on consultation and a detailed impact assessment.
The text also includes a reference to the exclusion of national security, defence and military purposes from the scope of AI legislation. Nor should the latter apply to AI systems and their results used for the sole purpose of research and development.
A further amendment was also made to specify that public entities will also be obliged to register in the EU database when using high risk AI systems.
Concerns about remote recognition
Following the announcement of the adoption of its position by the Council of the EU, some, such as German MEP Patrick Breyer (Greens/EFA), expressed their concerns about some of the provisions of the text adopted by the Council of the EU, notably biometric recognition.
“We must defend a society of trust and rights, not a society of suspicion and division. Mass surveillance has no place in our society and we will fight for it to be banned in the European Parliament” he said.
Several associations also issued a statement, sometimes regretting, like the European Consumer Organisation (BEUC), the adoption of “a disappointing position”, leaving “too many important questions unanswered, such as facial recognition by private companies in places accessible to the public”.
“It is crucial that Parliament stands up for consumers when it adopts its position on AI legislation in the new year. Otherwise, this legislation could end up being a disappointment”, commented BEUC Deputy Director General Ursula Pachl. (Original version in French by Thomas Mangin)