Members of the European Parliament’s Special Committee on Artificial Intelligence (AI) in the Digital Age adopted the draft report by Axel Voss (EPP, Germany) on AI on Tuesday 22 March (25 votes in favour, two against, six abstentions). The document proposes - in addition to a state of play - a roadmap to 2030 to close the EU’s gap in this area while protecting fundamental rights and democracies.
“The EU now has a unique opportunity to promote a human-centred and trustworthy approach to AI, based on fundamental rights, that manages the risks while fully reaping the benefits that AI can bring to society as a whole”, said Axel Voss.
Firstly, from a technical point of view, the German MEP’s report calls on the Commission to propose only legislative acts in the form of regulations on future digital legislation, so as to promote global harmonisation.
The report’s recommendations also call for the creation of a mechanism with “sufficient resources” to oversee the implementation of future AI legislation.
On this point, the report prefers the idea of the creation of a European AI Board to that of the creation of “an expensive new European agency for AI”.
This Board would be composed of the national AI supervisory authorities as well as several European bodies such as the EU Agency for Fundamental Rights , the High Level Expert Group on AI or the EU Agency for Cybersecurity.
Monitoring and ‘high risk’ uses
The text also notes the importance of basing the EU’s approach on people and states that co-legislators should, in future negotiations, “strive to align the definition of AI” with those of “other like-minded OECD democratic countries” (see EUROPE 12915/14).
Furthermore, Axel Voss’ report argues that the EU should not always regulate AI as a technology, but rather its level of regulatory intervention should be proportionate to the type of risk associated with using an AI system in a particular way.
Furthermore, “the classification of technologies as ‘high risk’ should be based on the actual use and context, the complexity and autonomy of the AI system, the probability and likelihood of the worst-case scenario, the severity of the harm and its irreversibility”, the paper says.
More broadly, the report also calls on Member States to rethink their approach, saying that national strategies - where they exist - “remain vague and lack clearly defined objectives”.
The draft report also brings up the issue - much debated in the European Parliament (see EUROPE 12912/35) - of mass surveillance and the ethical and legal issues that AI could raise.
“These rules must be in line with our values: democracy, the rule of law, fundamental rights and respect for the rule-based international order. This is crucial, as the struggle between authoritarianism and democracy is becoming more and more acute”, stressed Dragoş Tudorache (Renew Europe, Romania).
Finally, the report considers that investment will be needed to catch up in a number of areas. In 2018, the paper points out, private investment in AI in the EU peaked at €3.4 billion. In the United States, it amounted to €31 billion, and €21 billion in China. It was the same issue with public investment, with €1 billion invested in the EU in 2018 compared to €6.8 billion in China and €5.1 billion in the US.
“The Union is currently on the losing side, far from fulfilling realising its aspiration to become a leading global player in AI. There is still a small window of opportunity to change this situation, even though it will close very soon”, the report by Axel Voss concludes.
See the report: https://aeur.eu/f/ww (Original version in French by Thomas Mangin)