EU law on artificial intelligence (AI) came into effect Programmers now need to pay attention to the AI Act

Related Vendors

Universitätsklinikum des Saarlandes

FAULHABER Drive Systems

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

On August 1, the EU law on artificial intelligence (AI) came into effect. This officially regulates what AI is allowed to do and what it is not allowed to do in the EU. But what does this mean for programming practice?

A team led by Professor of Computer Science Holger Hermanns from Saarland University and Law Professor Anne Lauber-Rönsberg from the Dresden University of Technology has now examined how the AI Act affects the practical work of programmers and published their findings in a paper that will be published in the fall. Here are some initial information:

It becomes serious rather with high-risk systems

"The AI Act shows that politicians understand that AI can also be a danger, especially when it intrudes into sensitive or health-relevant areas," comments Hermanns. But how does the AI Act actually affect those who create it? "What do I need to read from that thing," Hermanns summarizes the question of countless affected parties. Because not every programmer will want to—and be able to—read through the (in German) 144-page regulation from beginning to end.

Together with his doctoral student Sarah Sterz, postdoctoral researcher Hanwei Zhang, and Anne Lauber-Rönsberg, a law professor at TU Dresden, and her research assistant Philip Meinel, Hermanns has written the aforementioned paper titled "AI Act for the Working Programmer", which fundamentally answers this question. The core of their insight is that, in the end, developers and users are not likely to notice much change. "The AI Act's regulations mainly become relevant when developing so-called high-risk systems," Sterz notes.

For example, AI in computer games is not affected

This is because the European AI law aims specifically at protecting the later users of a system from an AI that could be harmful or unfair to them. If an AI does not intervene in sensitive areas, it is not subject to the extensive regulations for high-risk systems, it is said. Hermanns underscores this with a concrete example: "The developers of an AI software, which scans applications and possibly filters out applicants before a human from the HR department has even looked at it, is subject to the AI Act's regulations as soon as the program hits the market or is put into operation." However, the AI that simulates the reactions of opponents in car racing games can continue to be developed and marketed without their developers having to worry about the AI Act, says Hermanns.