The independent High Level Expert Group on AI is pursuing its mission. After publishing its ethic guidelines in April, it launched a pilot phase on Wednesday 26 June to evaluate the list of criteria made available to companies. It has also published its policy and investment recommendations for the Commission and Member States.
As detailed in our columns earlier this week, they advocate risk-based AI governance, including a comprehensive mapping of relevant European laws (see EUROPE 12281/13). “We agree that there is no need for premature regulation that could stop innovation. That is why we proposed this approach [in stages: Editor's note]”, said the Commissioner for the Economy and the Digital Society, Mariya Gabriel, in the press room.
The red lines of the Expert Group
In its 50-page document, the Expert Group opposes mass surveillance or marketing, lethal weapons and a rating system to assess citizens. This group - which includes engineers, philosophers, lawyers, industrialists and a trade unionist - stresses that artificial intelligence must serve the health and safety of workers only to replace dangerous tasks. It also proposes to require those deploying AI systems to develop a self-identification system so that users are informed that they are not interacting with a human.
Guidelines on the ethical dimension
These recommendations are in addition to a “first deliverable”, as the experts call it: the guidelines on the ethical dimension of artificial intelligence. These guidelines, published on 8 April, establish key principles for a trustworthy AI, provide guidance to designers and provide a 'concrete and non-exhaustive' assessment list to achieve this trust (see EUROPE 12228/9, 12231/10).
It is on this third component - the evaluation list - that the piloting exercise that has just been launched will focus on. Andrea Renda, researcher at the Centre for European Policy Studies (CEPS) and member of the Expert Group, said that this process would be structured around two components: one in the form of an open-ended questionnaire, published today, the other in the form of a detailed analysis, with the experts of the High Level Group planning to go beyond their limits to analyse, over a day and in different sectors, the relevance of these criteria.
“I invite you to participate because, afterwards, we will conduct an analysis to decide which evaluation criteria should remain, be strengthened or modified to reflect reality as much as possible”, the Commissioner said during the European AI Alliance meeting.
Reactions
Overall, reactions to the work of the expert group were rather positive. The digital technology industry DigitalEurope let it be suggested that it did not agree with all the recommendations, but welcomed the “inclusive and agile” work process chosen to address these issues. The European Telecommunications Network Operators’ Association ETNO considered that these recommendations defined the right level of ambition, particularly welcoming the risk-based approach and mapping exercise recommended by the experts.
Thiébaut Weber, who participated in the Expert Group on behalf of the European Trade Union Confederation (ETUC), said: “This report contains clear recommendations for AI to serve humans, including in the workplace (...). The ball is in the court of the European Commission, the States, but also for employers and trade unions to seize these solutions”. (Original version in French by Sophie Petitjean)