The long-awaited ‘guardrail’ for AI, the European rules for artificial intelligence, was endorsed this week by the Council of EU Ministers. The rules, however, will not come into effect until 2025, while it was intended that his would happen (at least partially) this year.
Purpose of the AI Act
This AI law is the first legal framework for AI, and addresses the risks of artificial intelligence. The aim of the rules is to promote trustworthy AI in Europe and around the world. AI systems are to respect the fundamental rights and security of citizens, as well as limit the risks of powerful AI technology.
The main goal is to prevent citizens from becoming victims of companies and governments that work with large amounts of data and algorithms.
Multiple Risk Levels
Forms of AI technology are assigned a certain risk level, varying from ‘Minimal risk’ to ‘Unacceptable risk’. These new laws around AI focus on all of these different risk levels. The first set of rules, coming into effect in the beginning of 2025, focuses on the most dangerous applications of AI.
These rules specifically target artificial intelligence that is built to exploit weaknesses or manipulate people. Techniques to track or direct individuals, as China does with its population for example, will also be banned.
Rules Generative AI
Part of this AI Act is a set of rules for generative AI, such as chatbots. OpenAI’s ChatGPT and Google Gemini are well-known examples of such generative AI. These rules will come into effect around the summer of 2025.
