The EU Parliament is about to hold a landmark vote on the world’s first comprehensive law to regulate artificial intelligence (AI). With the European AI Regulation, the EU wants to create a legal framework for the development and use of AI. Similar to the Data Protection Regulation, the EU Commission hopes that other countries will follow suit. The regulation was the result of an 18-month balancing act by legislators in Brussels.
A central aspect of the regulation is the classification of AI according to the risks of its application. AI is divided into four categories: low-risk, limited-risk, risky and prohibited AI. Prohibited AI includes, for example, “social scoring” systems that evaluate people’s behaviour, automated recognition of emotions and blanket surveillance with real-time biometric data in public.
However, there are still open points of contention, especially regarding the scope of the use of biometric data in law enforcement. Some members of the Christian Democratic EPP Group are calling for changes in the draft to allow exceptions for the search for missing children and the fight against terrorism. At the same time, the Greens, the Left and the Pirates warn against mass surveillance.
With regard to the use of AI in different applications, a differentiated view is taken. High-risk AI, such as autonomous vehicles, is to be banned, while AI-powered toys and generative AI, such as the chatbot ChatGPT, remain permitted in principle. The higher the risk of the AI application, the stricter the requirements must be met. Manufacturers must assess the risks of their products and comply with certain standards for training data. Testing authorities are to control these measures.
Another important aspect of the regulation concerns the prevention of discrimination through AI. It is stipulated that AI must not make use of biased data sets, for example, to assess people’s creditworthiness or to be used in employee acquisition.
Expectations and fears about the European AI law are high. While Sam Altman, CEO of OpenAI, warns of the existential dangers of unregulated AI, smaller developers fear difficulties in implementing the extensive documentation requirements. Civil rights activists hope that the AI law will counteract discrimination and disadvantages.
After the upcoming vote in the EU Parliament, negotiations with the EU member states will begin. Especially the regulations on the use of AI in law enforcement are likely to cause further discussions. An agreement before the European elections next year would probably allow the AI regulation to come into force in 2026. It is foreseeable that further AI regulations will follow in the European Union as AI continues to evolve.