WS
AI Act1 min czas czytania1006 znaków137 słów
The AI Act — the first deadlines already apply.
The AI Act comes in stages, and some of its provisions already apply. If you build or use AI systems, it’s worth checking which risk category your product falls into.
The AI Act comes in stages, and some of its provisions already apply. If you build or use AI systems, it’s worth checking which risk category your product falls into.
The Artificial Intelligence Act (AI Act) introduces a risk-based approach — from prohibited practices, through high-risk systems, to transparency obligations for the rest. Importantly, the provisions come in stages, not all at once.
Alert regulacyjny — AI Act · EU · first deadlines apply
Prohibited AI practices and some obligations for general-purpose models have come into force in stages. Further deadlines are spread over time — they must be tracked for the specific product. As of May 2026.
Where to start
The first step isn’t implementing requirements but classification: whether your system is an "AI system" within the meaning of the regulation at all, and if so — which risk category it belongs to.
- Classify the system against the definition and risk categories
- Check whether prohibited practices apply to you
- Establish transparency obligations toward users
- Prepare documentation appropriate to the category
Most products won’t fall into the high-risk category. But even then, transparency requirements may apply — and ignoring them often costs more than the adjustment itself.
