The regulation applies step by step. The requirement for AI literacy among staff has applied since February 2025. The transparency requirements, for example that a chatbot must say it is an AI, have applied since August 2026. The requirements for high-risk systems start to apply in December 2027 for standalone systems and in August 2028 for AI in regulated products.
The first step is to classify your AI systems. Most of them, such as an internal assistant or a chatbot, end up in the lower risk classes with few requirements. Systems that affect decisions on employment, credit or access to public services may count as high risk.
The risk classification is part of our AI Discovery Sprint, so that you know which requirements apply before anything is built. Always check the dates against the current text, since the regulation has changed since it was adopted.