In Brussels today, the European Commission unveiled a much-anticipated new rulebook on Artificial Intelligence.

The Artificial Intelligence (AI) package, which comprises:

This first-ever set of rules on AI follows on from the AI Guidelines published by a High-Level Expert Group in 2019 and the Commission’s AI White Paper published in 2020.

The Commission’s ambition is to make the European Union into a global leader in the development of secure and trustworthy AI. At the launch of the package in Brussels on 21 April, European Commission Executive-Vice President, Margrethe Vestager, stated that an “ecosystem of trust goes hand in hand with an ecosystem of excellence”. The proposal aims on one hand to build trust in AI systems to mitigate associated risks and on the other hand to boost investment and innovation in the further development of AI.

With these goals in mind, the Commission has opted for what it regards as a proportional and risk-based approach, based on the following key assumption: “the higher the risk, the stricter the rules”. The object of the regulation is then not AI per se, but how AI is used. On this basis, AI uses have been classified in four different categories (minimal risks, limited risks, high risks and unacceptable risks). AI systems deemed to be “high risk” are a key focus in the proposed regulation and will be subject to a set of five specific obligations. Remote biometric identification systems, for example, fall within the “high-risk” category. The Commission also identified “unacceptable” AI systems, on which it proposes to place a ban as they are considered contrary to EU fundamental rights. With this proposal, the Commission has underlined that the evolution of European AI should be anchored to EU fundamental values or, in the words of EVP Vestager, to “what Europe is and what it stands for”. At the same time, the Commission aims to equip businesses with a set of rules to follow when it comes to the design and development of new AI systems. As the EU Internal Market Commissioner, Thierry Breton, commented: “if you want to make AI now, come to Europe. You will know how to do it”.

Next Steps

The European Commission’s regulatory package on AI will now need to be approved by both the European Parliament and Member States in the European Council, which will take a minimum of 18 months. This publication of the proposals therefore marks the beginning of what is expected to be an intense negotiation at EU level. During the next months, there will also be opportunities for industry stakeholders to contribute to the decision-making process.