The European Parliament has passed what it claims is the world's first comprehensive AI law. It is intended to make the use of technology safer, for example with requirements for facial recognition or ChatGPT.
The EU is once again hoping for the Brussels effect: it wants to set standards that the rest of the world follows. This is what happened with data protection and chemicals. Now Europe wants to become a pioneer in the regulation of artificial intelligence.
Technical systems that imitate human learning are developing extremely quickly – with all the opportunities and risks. The EU will regulate this in the future with an AI law. The EU Parliament and member states agreed on the draft in December. Now the plenary session in Strasbourg has given the green light with a large majority. 523 MPs voted in favor, 46 against and 49 abstained.
Law should regulate artificial intelligence more closely
The responsible commissioner, Thierry Breton, described the new rules as “historic”. According to him, this is the first law in the world to regulate artificial intelligence in a “balanced manner”. It would also “protect against excesses through misuse” but would also promote innovation.
In a debate that has lasted for years, the EU has struggled to find a balance: facial recognition is useful for unlocking cell phones, but it is dangerous when it is used to fish photos from the Internet that end up in dubious databases. The law is intended to protect against possible dangers, including those caused by deceptively fake images or audio files. At the same time, the EU wants to promote the enormous economic opportunities of AI.
“Bureaucracy and ambiguity remain”
Not all MPs are completely satisfied with the balance. The Christian Democrat Axel Voss explains: “We shouldn't think that our work ends with this law. No, I think it's just beginning. Unlike other laws, we have to find adjustments to problems and developments more quickly.”
The law basically divides AI systems with high risk, to which special requirements apply, and applications with low risk potential with few requirements. Users must be able to recognize that content is AI-generated.
FDP MP Svenja Hahn fears that the requirements could cause problems, especially for smaller companies – so-called SMEs: “But bureaucracy and ambiguities remain, which are easier for BigTech companies to cope with than for SMEs. This is where it becomes important that the Commission creates clarity in implementation.” The law should not become a brake on innovation.
security authorities are allowed to continue face recognition to use
Systems that divide people into categories based on their skin color, religion or sexual orientation are prohibited. The untargeted reading of images from the Internet or from surveillance cameras for databases is also not permitted. There are exceptions for biometric identification in public spaces in real time, for example when the police want to prevent an attack or are looking for victims of human trafficking.
This goes too far for the left. Sergey Lagodinsky from the Greens admits: “This law is weaker than we wanted in parliament. But it is much better than having no regulation at all, than conditions like the Wild West with AI.”
The vote from the EU states, which is expected at the end of April, is still missing. However, this is considered a formality. It will take another two years for the law to fully apply.
Jakob Mayr, ARD Brussels, currently Strasbourg, tagesschau, March 13, 2024 1:40 p.m
