European actions on AI regulation
Explore the comprehensive overview of the EU's actions on the AI Act regulation, from its adoption to the latest developments.
We will ensure that AI will be “the best thing ever to happen to humanity” (Speech by Commissioner McGrath on How AI Governance shapes Europe's Geopolitical Position, May 14th, 2026)
The EU AI Act: Regulation in motion
The EU AI Act (Regulation (EU) 2024/1689) is the world's first legal framework meant to regulate artificial intelligence. It is built on a "risk-based approach", meaning that it classifies AI systems into four categories :
- Unacceptable risk : systems are completely prohibited. Since February 2025, eight categories of practices have been prohibited. This includes manipulation, the exploitation of vulnerable groups and policing based on profiling.
- High risk : systems must undergo strict obligations before being placed on the market. According to the European Commission, approximately 5 to 15% of AI systems on the EU market fall into this high-risk category. TThis requires a mandatory conformity assessment. and examples include recruitment tools, credit systems, medical devices and critical infrastructure.
- Limited risk : systems are subject to transparency requirements ( chatbots or deepfakes) so that users are informed they are interacting with AI
- Minimal risk : the default category covering the vast majority of applications (spam filters, video games), without specific regulatory requirements.
It imposes obligations proportional to potential harm.
It is therefore a pioneering law, it impacts thousands of businesses developing AI within the EU.
A phased implementation
Recognizing the need for industry adaptation, the EU opted for a staged rollout of the AI Act. This approach allows businesses and developers sufficient time to comply with the new regulations.
February 2, 2025
Phase 1: The initial phase of implementation commenced on February 2, 2025. During this phase, AI systems identified as posing an unacceptable risk became subject to the regulation's requirements.
Economic effects and developments in 2026
The EU AI Act is evolvingn, the phases in 2026 introduce obligations and guidelines for different categories of AI systems.
The heaviest fines, up to €35 million or 7% of annual global revenue are reserved for violations involving prohibited systems. If there is no compliance with the obligations applicable to high-risk systems ut exposes companies to fines of up to €15 million or 3% of global revenue.
Also, providing incorrect or misleading information to the authorities can result in penalties of up to €7.5 million or 1% of annual turnover.
Thus, the economic impact is considerable. Compliance with a single high-risk AI system costs an average of around €52,000 per year, and annual compliance expenditure averages €29,277 per system for affected companies. The market for AI regulatory compliance could reach between €17 billion and €38 billion by 2030.
This is therefore a pioneering law with global reach because its scope is extraterritorial. This means that any company, even one based outside the EU (in the United States or Asia), is subject to its rules as soon as its AI systems have an impact on users established within the Union.
Create Your Own Website With Webador