EU AI Act in plain English
What the AI Act actually requires — and when each piece bites
The EU AI Act is now in force and applies in stages through 2026 and 2027. It classifies AI systems into four risk tiers: prohibited (e.g. social scoring), high-risk (e.g. AI in employment, credit, education, critical infrastructure), limited-risk (chatbots, generative AI — transparency obligations) and minimal risk (most everything else).
If any of your AI use cases falls into 'high-risk', the obligations are substantial: a quality management system, data governance, technical documentation, record-keeping, transparency to users, human oversight, accuracy and robustness, cybersecurity, conformity assessment and registration. Most B2B AI products land in 'limited risk' — which still requires user-facing disclosures and clear labelling for AI-generated content.
- Prohibited AI: in force from February 2025.
- General-purpose AI model obligations: from August 2025.
- High-risk AI obligations: phased through August 2026 / 2027.
- Penalties: up to €35m or 7% of global turnover for prohibited-use violations.
