The European Union has adopted uniform rules for the use of artificial intelligence – and hopes to set a standard worldwide. However, critics fear that the regulations could quickly become outdated.
The EU states have decided on stricter rules for artificial intelligence (AI) in the European Union. In Brussels they agreed to plans that regulate the use of the technologies, for example in video surveillance, speech recognition or the analysis of financial data. According to EU countries, the law, which will only come into force in spring 2026, is the first law of its kind in the world. It could set a global standard for the regulation of AI.
The regulation aims to make the use of AI in the EU safer. It is intended to ensure that AI systems are as transparent, comprehensible, non-discriminatory and environmentally friendly as possible. An important aspect is that the AI systems are monitored by people and not just by other technologies. The plans go back to a proposal from the EU Commission in 2021.
The law applies to anyone who develops, offers or uses AI systems within the EU. This affects public and private actors both inside and outside the EU.
Strict requirements for “high-risk” applications
Among other things, the regulation provides for a labeling requirement. Developers should have to mark texts, sounds and images generated with artificial intelligence in order not to mislead people. However, experts consider this to be difficult to control because of the abundance of material.
Systems that are considered particularly risky and are used, for example, in critical infrastructure or in education and healthcare, will also have to meet strict requirements in the future. This includes facial recognition at train stations or other public places. A court order is now required. Police and other security authorities should only be allowed to use such facial recognition in public spaces to prosecute specific crimes such as human trafficking or terrorism.
Certain AI applications that violate EU values should be banned entirely. This includes, for example, mass surveillance with biometric data or the assessment of social behavior (“social scoring”). In China, for example, citizens are divided into behavioral categories.
Digital association criticizes delayed regulation
In light of the recent reform of EU asylum law, Amnesty International had warned against using facial recognition and other controversial technologies against migrants and other people seeking protection. European MPs from the SPD and the Left also criticized the police’s identification of people in real time as not being regulated strictly enough.
The digital association Bitkom criticized the fact that the AI law leaves essential questions unanswered. In Germany and the other EU countries, the regulatory work is only now beginning, said association president Ralf Wintergerst. Whether AI receives a boost in Germany and Europe or, above all, is faced with new obstacles depends crucially on how this framework is designed and the regulations are implemented in Germany. Other critical voices also fear that the EU requirements could be outdated again in two years.
Wissing for “innovation-friendly” implementation
Federal Digital Minister Volker Wissing admitted that the legislature is constantly challenged with a technology that is changing so quickly such as AI. “We cannot expect that we will finally clarify future questions with regulation,” said the FDP politician. “That’s why I’ve always been in favor of us getting on the path to regulation quickly, but also having the courage to continually make adjustments.” It is important that the innovation-friendliness of the regulation is always kept in mind.
According to its own statements, the federal government had worked at the EU level to ensure that so-called general-purpose AI such as the chatbot ChatGPT should not be classified as a high-risk application.
Violations result in fines
Following confirmation from EU countries, the new rules will now be published in the Official Journal and will come into force 20 days later. They should then apply two years after they come into force. Until the law takes effect in spring 2026, the European Commission will rely on voluntary agreements with companies.
If they do not comply with the rules after the law comes into force, member states must decide on sanctions. These can be fines. Private individuals who discover violations of the rules can complain to national authorities. These can then initiate monitoring procedures and impose penalties if necessary.