When you're sitting at the breakfast table, taking your first sip of espresso, your body still doesn't quite know whether to get ready for work or mowing the lawn - and then you stumble across the term "EU AI Act". In other words, a law. Sounds like a very dry theatrical performance in which paragraphs stumble over paragraphs to explain to each other how dangerous they are. But far from it: this law is so much more. It is drama, opera, sometimes even a cabaret of bureaucracy. And yet it affects us all.
It began as so many things do: with good intentions. In 2021, the European Commission thought it would be nice if artificial intelligence - which at the time was already capable of more than a mid-level apprentice in their third year - was somehow regulated. There was much discussion, honing, praise, scolding and revision. Until finally, on July 12, 2024, the final work appeared in the Official Journal of the European Union. A monumental text. A regulatory monolith. And yes - also a bit of a masterpiece of paragraph poetry.
However, the actual effect did not take effect until August 1, 2024. In force, but not with a thunderclap. Rather, the EU AI Act crept into the legal system like a cat into an unobserved living room - first curiously, then permanently.
| Date | Regulation |
|---|---|
| Aug 1, 2024 | Entry into force |
| Feb 2, 2025 | Prohibition of certain unacceptable risks, AI literacy obligation |
| Aug 2, 2025 | Obligations for GPAI models, enforcement by AI Office, sanctions & authority structure specified |
| 2 Aug 2026 | Regulation fully in force (start of transition periods for high-risk systems) |
| 2 Aug 2027 | Obligations for high-risk AI are binding |
This staggered approach allows companies not to rush headlong into a panic sweat. No, you can make a plan. And tea. Lots of tea.
The first stage of the law is already having an impact: unacceptable risks, such as manipulative systems, social scoring or biometric real-time recognition in public spaces, are strictly prohibited. Emotion recognition, on the other hand, remains subject to differentiated regulation: It is prohibited in sensitive contexts such as in the workplace or in educational institutions for performance evaluation. In other contexts, such as market research (e.g. at trade fairs), it remains permissible provided that it is transparent.
Another new aspect is AI literacy. Employees need to learn what they are actually using. A step that was as logical as it was overdue.
From August 2025, extensive obligations will apply to general purpose AI models (GPAI). Transparency, disclosure of data sources, copyright issues and documentation requirements will take center stage. In addition, the central EU authority AI Office, which has been active since February 2024 and has been working on standards and guidelines since then, will provide the first enforcement mechanisms. The member states must also appoint their national supervisory authorities, which will be responsible for monitoring.
For the first time, sanctions, new confidentiality requirements and the new roles of notifying bodies for compliance checks will also apply. This marks the start of the operational phase of the governance structure of the EU AI Act.
Now you might be inclined to lean back with a mixture of resignation and chamomile tea. However, as strict as the EU AI Act may seem, it is not just a warning, but also an invitation. An invitation to integrate AI into business practice in a responsible, transparent and future-proof manner. Those who check the new requirements at an early stage, categorize their systems and document processes properly will not only gain regulatory certainty, but often also a real competitive advantage. Compliance becomes a business card, an invitation card for new customers and partners who are looking for trust in a world full of algorithms and uncertainties.
Here, so to speak, in the middle of the engine room of this new AI world. We guide companies through the pitfalls and footnotes of the EU AI Act. We translate paragraphs into practice, create structures, clarify responsibilities and help with the risk-based classification of your AI systems. Between AI literacy, GPAI obligations and the subtle nuances of transparency rules, you need someone who not only knows the theory, but also understands the operational challenges of everyday life.
And if one morning, as you take your first sip of espresso, that term dances through your head again - the EU AI Act - then you can sit back and relax. We have that in mind.
