The European Commission does not sufficiently understand the need for a better AI Act
The Dutch Senate shares the concerns we have about the Artificial Intelligence Act and wrote a letter to the European Commission about it. The Commission has given a response to this which is not exactly reassuring.
The gaps in the AI Act
At the end of last year, the Senate expressed its concerns to the European Commission about the Artificial Intelligence Act. In a letter, the Commission was questioned about the formulated prohibited applications, including social scoring, the abuse of vulnerabilities of groups of people and the use of biometric identification. Concerns have also been expressed about the risk-based approach and legal protection. We are pleased to read that the Senate has voiced the same concerns we’ve expressed. Unfortunately, we are less happy with the answers from the European Commission.
Leaving the job half done
While more and more institutions and organisations plead for a complete ban on biometric surveillance in publicly accessible spaces, the European Commission maintains its position that the ban is limited to real-time biometric identification. These are only uses of artificial intelligence that can recognise people instantly or at least “without considerable delay”. This still allows all kinds of applications. Including applications where identification takes place afterwards that can just as easily be used to follow people in public space. There are even a number of risks associated with post biometric identification systems that are not present with real-time applications. For example, the images can be tampered with. Or large quantities of image material can be examined afterwards to see whether someone appears in them and where someone has been and with whom. This entails a great risk for whistle-blowers or the source protection of journalists.
Contradictory answers
On the one hand, the European Commission seems to want to reassure the Dutch Senate by indicating that there is already a ban on the use of biometric data in European privacy law. This includes an exception for law enforcement purposes. Some countries in the European Union have legislation to specify that exception, others have not. That is why the European Commission now finds it necessary to ensure a ‘level playing field’ by providing the same rules throughout the European Union. There is, of course, something to be said for this.
What is worrying, however, is that a few paragraphs later the European Commission lists all kinds of ‘advantages’ of biometric identification that are questionable, to say the least. For example, recognising emotions could be useful for blind people and people with an autism spectrum disorder. It could warn people about falling asleep at the wheel. And it could check the age of gaming children to protect them from harmful content.
The idea is all very noble, but as yet it appears that this type of applications are full of errors. The faces of people of color are poorly or not at all recognised and there is no scientific basis for being able to derive emotions from the external features of a face. Moreover, it is based on the assumption that all people express emotions in the same way, and facial expressions always mean the same thing, while this is not the case. If only because as people from different cultures and environments we express our emotions differently. And is the use of biometric identification really what we want to use to protect our children?
It’s all very noble thinking, but as yet it appears that this type of applications are full of errors.
It is not reassuring that in listing what it sees as advantages, the European Commission completely ignores the already known risks to human rights associated with these applications.
In the coming period, the European Union will discuss the legislative proposal in Parliament and in the Council. We remain committed to achieving a law that effectively protects fundamental rights.