Anne Meuwese on EU's impending AI regulation
This regulation – also known as the AI Act – aims to ensure that AI systems sold and used in the EU are safe and consistent with existing fundamental rights legislation and Union values. AI harvests its factual material on the Internet, but in some cases it can be misleading. This is sufficient reason for the European Union to impose rules to protect its citizens from the dangers of smart software.
Europa eager to set standard
The European Union now leads the world when it comes to developing regulations, observes Professor of Public Law Anne Meuwese of Leiden University. 'Either you let Big Tech determine the playing field, or you take the initiative yourself. Europe is eager to set the standard. European rules will then hopefully be followed by other countries.'
The European Commission has pulled out a ‘very standard toolbox’ for this purpose, Meuwese says. ‘European regulation of the internal market and product safety already exists, with fairly broad powers granted by the legislature. It has been decided that artificial intelligence can also be seen as a product. It will have to meet certain conditions before it can be offered on the European market, just like with cars and airplanes.' The greater the risks inherent in a system, the stricter the regulation.
Many questions remain
At the same time, many questions remain. AI products are constantly being updated – how does that relate to the permission required to market them? Is that always a new product, with possibly a different risk profile? If there are substantial changes, you have to reapply for approval. ‘But what counts as a substantial change?’ wonders Meuwese. And what if a user modifies such a product himself? 'How is it determined whether you then have to go through an approval process again?' Although Meuwese is positive about the new rules, she sees that many issues remain to be resolved.
Respect for human rights
'The Council has chosen a totally different approach’, says Professor Meuwese. 'It wants to provide protection against specific human rights violations arising from the use of AI, such as discrimination. That approach places far more obligations on individual countries themselves. They have to safeguard respect for human rights and will have to deal with violations.' From the citizen's perspective, this is a smart approach according to Meuwese. 'What can you do, for instance, if you are discriminated against by your municipality as a result of an AI system being used? If that system meets European requirements, it is difficult to bring a case under the European Union's product-oriented AI regulation. Going to the administrative courts to claim discrimination is then more obvious. I expect that on the basis of that treaty, citizens will find it much easier to obtain justice.'