The European Union’s AI Act officially entered its enforcement phase on February 2, 2025, making it the world’s first comprehensive AI regulation framework. Companies using high-risk AI systems now face compliance deadlines ranging from 6 to 36 months, with penalties up to €35 million or 7% of global revenue for violations.

What AI Systems Are Classified as High-Risk Under These Regulations?

The EU AI Act identifies eight categories of high-risk AI applications, including biometric identification systems, critical infrastructure management, employment decision tools, and credit scoring algorithms. AI systems used in education, law enforcement, and migration management also fall under strict oversight. The European Commission estimates these rules affect approximately 15% of AI applications currently deployed across member states.

When Do Companies Need to Comply With the New AI Regulations?

Implementation follows a staggered timeline: bans on prohibited AI practices took effect immediately in February 2025, general-purpose AI model requirements begin August 2025, and high-risk system compliance is mandatory by February 2027. The European AI Office will coordinate enforcement across all 27 member states.

How Much Will Non-Compliance Cost Businesses?

Penalty tiers range from €7.5 million for incorrect information submission to €35 million for deploying prohibited AI systems. These fines can alternatively be calculated as 1.5% to 7% of worldwide annual turnover, whichever amount is higher—making this potentially more expensive than GDPR violations.