Allaying fears and avoiding drift while maintaining the course. With these objectives in mind, the Executive Vice-President for a Europe fit for the Digital Age, Margrethe Vestager, and the European Commissioner for Internal Market, Thierry Breton, presented the European Commission’s proposal for a new regulation on artificial intelligence (AI) on Wednesday 21 April.
In this balancing act, the Commission emphasised the trust of citizens, at a time when more and more questions are being raised in civil society, notably concerning remote biometric identification and the reliability of autonomous vehicles.
“On Artificial Intelligence, trust is a must, not a nice to have. With these landmark rules, the EU is spearheading the development of new global norms to make sure AI can be trusted. By setting the standards, we can pave the way to ethical technology worldwide and ensure that the EU remains competitive along the way”, Ms Vestager insisted.
“The Commission has presented an ambitious and far-reaching proposal. It is clearly, and unequivocally, focused on the protection of European citizens and their rights and provides much-needed initial clarity on what is and is not allowed in artificial intelligence”, Dragoș Tudorache (Renew Europe, Romania), chairman of the European Parliament’s special committee on Artificial Intelligence in the Digital Age (AIDA), agreed in a statement.
Thus, the Commission’s proposal focuses more on the regulation of AI uses than on the regulation of the technology itself. For, as several senior European officials have hammered home, this would not mean forgetting the Commission’s other desire, namely to make the EU a champion of artificial intelligence.
“AI is a means, not an end. [...] It offers enormous potential in sectors as diverse as health, transport, energy, agriculture, tourism and cybersecurity”, said Mr Breton, before continuing: “This also entails a number of risks. Today’s proposals seek to consolidate Europe’s position as a global centre of excellence in AI, from the lab to the market, to ensure that AI in Europe respects our values and rules, and to exploit its potential for industrial use”.
In concrete terms, the legislative proposal follows a risk-based approach and is based on four categories. The first is for low-risk uses—which the majority of AI systems fall under—that will not be subject to third-party intervention. Uses representing limited risks will have to comply with transparency rules, such as specifying to users when they are interacting online with a chatbot.
Thirdly, AI systems considered “high risk”—in critical transport infrastructure, education, training, product safety or in the field of employment and private services—will have to comply with strict requirements in order to be put on the market and human control will have to be exercised. In addition to data quality, full transparency and detailed and accessible documentation of systems will be required.
Finally, the proposed regulation provides for an outright ban on systems that represent an unacceptable risk. This includes all uses of AI that threaten people’s security, livelihoods and rights.
Remote biometric identification at the heart of the debate
Exceptions will be made for remote biometric identification systems, labelled as “high risk”, when it is necessary to prevent a specific and imminent terrorist threat, to identify or locate a suspect in a serious crime or to search for a missing child.
However, the Commission states that the use of these systems will have to be authorised by a judicial body or an independent body. And this will be subject to limitations in terms of duration, geographical scope and databases consulted.
Unlike the Renew Europe group, which welcomes the legislative proposal, this enthusiasm is not shared by the whole of the European Parliament. “The European Commission’s proposal would bring the high-risk use of automatic facial recognition in public spaces to the entire European Union. The proposed procedural requirements are a mere smokescreen”, said Patrick Breyer, German Pirate Party MEP.
Fearing the advent of a European ‘Big Brother’, the S&D group welcomed the basics of the proposal, but warned that it would do everything to “defend fundamental rights, such as privacy in particular, when it comes to facial recognition in public spaces”.
In addition to remote biometric identification, several organisations have also expressed their disappointment, such as the European Consumers’ Organisation (BEUC), that the proposed rules provide too little protection for consumers. And the collective network of NGOs, experts and researchers European Digital Rights (EDRi) considered that the EU should go further to prevent surveillance and discrimination. “Whilst it is positive that the Commission acknowledges that some uses of AI are simply unacceptable and need to be prohibited, the draft law does not prohibit the full extent of unacceptable uses of AI. This leaves a worrying gap for discriminatory and surveillance technologies used by governments and companies”, detailed Sarah Chander, an AI expert for EDRi.
Another part of the digital package is the EU’s approach to new equipment and machinery. In this area, the Commission wants to ensure that the new generation of machines offers the necessary safety to users and consumers, while encouraging innovation. This should lead to greater legal clarity, reducing the administrative burden and costs for businesses by allowing documentation in digital format and adjusting conformity assessment fees for SMEs, the Commission said.
The coordinated plan foresees the use of funds allocated under the Digital Europe and Horizon Europe programmes (see EUROPE 12678/17), amounting to €1 billion each year. It should also be recalled that 20% of the expenditure of the Recovery and Resilience Facility, the budgetary instrument at the heart of the Next Generation EU Recovery Plan, should be devoted to the digital transition (see EUROPE 12561/7).
The Commission intends to rely on the national market surveillance authorities to carry out this task. The latter will have to ensure compliance with the new rules and will themselves be able to rely on the future European Artificial Intelligence Committee. This committee will also be responsible for stimulating the development of new standards for AI.
The proposals, the result of 3 years’ work—the publication of the European AI strategy dates from 2018 (see EUROPE 12009/4)—will have to be adopted in co-decision by the European Parliament and the Member States.
See the Commission’s proposal for a regulation: https://bit.ly/32BK0Lw (Original version in French by Thomas Mangin)