Artificial intelligence: AI Act may come into force
After the EU Parliament, the EU member states have approved the AI Act. Now it is important to quickly clarify some open questions, says Bitkom.
(Image: Anterovium/Shutterstock.com)
The European Council has approved the AI Act previously adopted by the EU Parliament. It can now enter into force as soon as it has been published in the EU Official Journal. The Act is intended to harmonize the legal provisions on dealing with artificial intelligence across the EU. The higher the risk of harm to society, the stricter the rules, as the EU member states explain in a statement. At the same time, the law is intended to create a basis for investments in safe and trustworthy AI systems.
AI systems with limited risk are to be subject to light transparency obligations, while high-risk AI systems will only be approved in the EU if they meet a number of stricter requirements. As examples, the EU Council lists systems that make social assessments or manipulate behavior. AI for preventive police work that uses profiling and biometric data to categorize people according to certain categories such as race, religion or sexual orientation is also prohibited.
In its communication, the EU Council pays particular attention to "general purpose AI" (GPAI), which also includes AI chatbots. Including these in the AI Act was not at all in the interests of companies such as Google or Microsoft. The Council now explains that GPAI models without systemic risks are subject to some limited requirements, for example with regard to transparency. Stricter rules should apply to GPAI models with systemic risks.
Four committees
The EU Commission is setting up an "AI Office" to concretize and enforce the new rules. A committee of independent scientific experts will support the office in its work. In addition, an AI Board with representatives from the EU member states will advise the Commission on how to apply the AI Act consistently and effectively. The fourth body will be an advisory forum consisting of representatives of interest groups who will support the EU Commission with their technical expertise.
Belgian State Secretary Mathieu Michel described the AI Act as groundbreaking. It is the first law of its kind in the world. It deals with a "global technological challenge" that also offers opportunities for societies and economies. The AI Act emphasizes the importance of trust, transparency and accountability in dealing with new technologies and at the same time ensures that they can flourish in Europe.
"Act quickly now"
This is very much in the spirit of the German IT industry association Bitkom. However, the AI Act still leaves key questions unanswered. "Whether AI receives a boost in Germany and Europe or is faced with new obstacles depends crucially on how this framework is designed and how the regulations are implemented in Germany," said Bitkom President Ralf Wintergerst.
The EU Commission must now set up the AI Office as quickly as possible and start implementing the requirements for GPAI models. "As these are due to come into effect as early as the middle of next year, it is particularly important that the regulation for these basic AI models is designed to be low-bureaucracy and practical," demands Wintergerst. The German government must quickly present a proposal for a national implementation law so that companies know what to expect.
(anw)
