Paul Nemitz: ‘We cannot leave AI to self-regulation and ethics alone’

Paul Nemitz is Principal Adviser in DG Justice and Consumer Protection of the European Commission and Professor of Law at the College of Europe. He is considered as one of Europe’s most respected experts on digital freedom and has led the work on the EU’s General Data Protection Regulation. The English translation of his essay The Human Imperative – Power, Democracy and Freedom in the age of Artificial Intelligence, co-authored with Matthias Pfeffer, will be published in June of this year. 

Voxeurop: Is Artificial Intelligence (AI)  an opportunity for democracy or a threat to it?

Paul Nemitz: Alarmist voices are increasingly replacing naive optimism about AI. Both Elon Musk and Bill Gates have said that AI is like nuclear power, an opportunity and an existential risk. In his book Human Compatible, Stuart Russel, author of the best-selling textbook on AI, describes the problem of controlling AI. He draws a parallel with nuclear power: We cannot be sure that general artificial intelligence will not be achieved by tomorrow. There was a time when all the leading scientists thought that splitting the atom was impossible. All this calls for democracy to take control of this technology, according to the time-tested precautionary principle. 

Responsible engineers and developers will not disagree. Neither the individual nor democracy can be controlled and manipulated by AI. On the contrary, individuals and democracy must remain in control of AI. Whether we can make AI an opportunity for democracy depends on many factors. First and foremost, the willingness of those developing AI to serve democracy and not just to make a profit. Second, a willingness to invest in AI specifically designed to empower democratic actors – from parliaments and governments, political parties, media, trade unions, NGOs and churches to individuals – to contribute more actively and constructively to the functioning of democracy.

Should the EU regulate AI, and if so, how and in what direction? Does the AI Act address the main issues related to AI and its use?

We cannot leave this technology to self-regulation and ethics alone. Like chemicals, cars and nuclear power, to name but a few examples, this technology is important enough to require a law to define its direction and limits. The AI Act will be an important precedent confirming the primacy of democracy in times of rapid technological development. The cheap calls for self-regulation and ethics rather than binding and enforceable law are outdated, because the power and speed of technological development simply require law to ensure that the public interest is served and that everyone, including those who do not want to play, is actually bound by rules that are enforceable. 

We cannot ignore the important issue of power when discussing AI. Without binding law, the power of technology to shape society lies solely in the hands of those who develop and own it. If society were organised in this way, democracy would not work, nor could we ensure respect for fundamental rights. The EU’s internal market also needs a regulation, because without a law at EU level, we would soon have a fragmentation of legislation across 27 Member States and thus no functioning internal market in high technology. The EU’s AI Act addresses many important issues related to the development and use of AI and, like everything in democracy, will be an act of compromise, a compromise in the right direction between different political world views.

Under the current draft legislation, AI tools will be categorised according to their perceived level of risk: from minimal to limited, high and unacceptable. Areas of concern could include biometric surveillance, spreading misinformation or discriminatory language. Does this make sense?

The risk-based approach to regulation is a good start. But it is limited because it reduces legislation to a repair shop for market failures and technological risks created in the private sector. So, if we only had risk-based regulation, democracy would be abandoning the aspiration that people, through democracy, shape their societies and the way they want to live. That said, the AI Act is part of a holistic package of first-generation legislation with which the EU is shaping the new digital realities. It stands alongside the Digital Services Act (DSA), the Digital Markets Act (DMA), the General Data Protection Regulation (GDPR) and consumer protection legislation, to name but a few pieces of legislation already in place. It now needs to be adopted quickly to create facts. 

The more than 3,000 amendments in the European Parliament show that democracy has a lot to say about AI and its regulation. And that it works well in Europe. I believe in the willingness to compromise in order to pass legislation and show that democracy can work. In this spirit, I believe the AI Act, together with other laws already in place that also apply to AI, is a good first piece of democratic legislation on Artificial intelligence, binding on both the private and public sectors. We can be proud that Europe is once again ahead of the game on this important issue of putting democracy before new technologies.

Can we trust the Silicon Valley tech giants engineers and moguls for self-regulating Ai? The recent call by a few of them to pause Ai’s development is going in the right direction or should governments (and the EU) intervene. Should there be a worldwide regulation, to h…