Below we report an article published in Milano Finanza, an Italian economic-financial daily, in which Sandro Gozi, Secretary General of the European Democratic Party, analyses Europe’s progress in pursuing world leadership in the field of reliable Artificial Intelligence.
Creating good laws is always difficult. Creating effective laws for activities that do not yet exist, in accordance with our model of society, becomes even more difficult. This is the challenge we decided to face in Europe with the new Artificial Intelligence law we adopted in Strasbourg. We have sought a good balance between two fundamental requirements of our strategy. On one hand, we want to promote and encourage research, innovation, the development of models, the promotion of open source software and fill our huge competitive gap with the United States or China. On the other hand, we do not want to innovate at the expense of our fundamental values and freedoms, which must be fully respected. Europe stands as a third way between American total laissez-faire and Chinese totalitarian control.
After our vote, we will start negotiations with the EU Council with the aim of reaching a final agreement by the end of the year. We aim to establish harmonised rules on AI, thus ensuring the proper functioning of the internal market and the free circulation of AI-based goods and services across borders. Regarding innovation, I believe that we will have to work further in the negotiation phase to favour even more flexible rules for our operators and businesses than the Parliament’s text, which is the result of a difficult compromise and which must evolve. Europe aspires to become a global reference point for a reliable Artificial Intelligence. With this law, we emphasise the need to ensure that AI is ethically responsible, technically robust and in accordance with the democratic values and fundamental rights of the European Union.
This is achieved first and foremost through a precise definition of artificial intelligence applicable to current and future systems. AI developers must respect principles such as human intervention and supervision, technical robustness, privacy protection and data governance, transparency, diversity, non-discrimination, fairness, and social and environmental welfare. However, we will have to make a greater distinction between model developers, who need to be given more margins, and application developers. This will undoubtedly be one of the priorities in the subsequent negotiations. We have also developed different risk categories: with a targeted approach to identify high-risk systems, while reducing bureaucracy for low-risk systems. Only systems that present a significant risk to health, safety or fundamental rights will be part of the former. We are pushing for a clear definition of the roles of the different actors in the AI value chain, with a focus on the needs of SMEs. For example, we are fighting against unfair contractual clauses imposed unilaterally. And above all, we want to promote innovation, e.g. by exempting research activities and freely licensed AI components.
New provisions on what we have called ‘sandboxes’ allow suppliers to test their AI systems without having to comply with the regulatory framework right away. However, even the most perfect law is of little use without an effective enforcement mechanism. We therefore call for the strengthening of the European PIA Office, which will be responsible for enforcing the rules. But this cannot come at the expense of weakening our values and freedoms. That is why we forbid invasive and discriminatory uses. This includes practices such as subliminal or manipulative deliberate techniques, exploitation of people’s vulnerabilities and use for social evaluation purposes.
The prohibition also extends to remote biometric identification systems, biometric categorisation based on sensitive characteristics, predictive policing systems, emotion recognition in various contexts and indiscriminate extraction of biometric data for facial recognition. Although most of these bans are widely supported, the absolute prohibition of real-time remote biometric identification systems in public spaces remains controversial. Some believe that there may be legitimate cases for the use of such systems, for instance in law enforcement, provided that proportionate and justified safeguards are in place. I think so too, and I believe we will have to work further on these aspects, which will undoubtedly be the focus of the forthcoming negotiations with the Member States.
Finally, the use of AI in politics raises growing concerns, especially regarding the rapid spread of negative content and fake-news. While the human brain could take hours to generate a thousand tweets of political criticism and insults, an AI could do it in less than 10 minutes. This ease with which PIA can spread negative content raises questions about the responsibility of online platforms. Facebook, in particular, has undergone major changes in recent months, largely due to the advent of AI. However, it remains to be seen whether other platforms have properly equipped themselves to deal with this issue, to preserve the freedom of expression while protecting themselves from abuse. All this requires the creation of an environment in which human users, identifiable and in the flesh, are free to express themselves, rather than being overwhelmed by trolls or AI software. We are taking an important step towards creating a solid regulatory framework for PIA in Europe and developing a model that is unique in the world. But we also need to work simultaneously on tools for investing in innovation. Having good rules is important, but being able to finance good investments is even better. This is the Europe we want: a multiplier of opportunities, freedoms and protections.
Sandro
Gozi
Secretary General
France
