EU AI Act › Risk Categories

The 4 Risk Categories of the EU AI Act

The EU AI Act classifies AI systems into four risk levels – from prohibited to minimal. The classification determines which obligations apply to your company.

The risk-based approach is the core of the EU AI Act. The higher the risk of an AI system, the stricter the requirements. Companies must correctly classify their AI systems to implement the right compliance measures.

Prohibited AI

Unacceptable risk

High-Risk AI

High risk

Limited Risk

Transparency obligations

Minimal Risk

Voluntary measures

Risk pyramid – from prohibited (top) to minimal (bottom)

Prohibited AI

Art. 5 EU AI Act

AI systems considered a threat to fundamental rights and safety are completely banned in the EU. These systems may not be developed or deployed.

Examples

Social scoring by public authorities

Real-time remote biometric identification in public spaces

Manipulation through subliminal techniques

Exploitation of vulnerabilities of specific groups

Obligations

Complete prohibition. No placing on market, no putting into service, no use allowed. Violations are punishable by up to €35M or 7% of annual turnover.

High-Risk AI

Art. 6–49, Annex III EU AI Act

High-risk AI systems affect critical areas such as health, safety or fundamental rights. The most comprehensive compliance requirements apply to them.

Examples

AI in medical devices and diagnostics

AI for HR decisions (recruiting, performance evaluation)

Credit scoring and insurance scoring

AI in critical infrastructure (energy, transport)

Obligations

Risk management system, technical documentation, data governance, conformity assessment, CE marking, registration in EU database, human oversight, logging and monitoring.

Limited Risk

Art. 50 EU AI Act

AI systems with limited risk are subject to transparency obligations. Users must know they are interacting with AI or that content is AI-generated.

Examples

Chatbots and conversational AI

AI-generated text, images or videos

Deepfakes and synthetic media

Emotion recognition in the workplace

Obligations

Labeling requirement: AI-generated or -manipulated content must be marked as such. Users must be informed when interacting with an AI system.

Minimal Risk

Art. 95 EU AI Act

The vast majority of AI systems fall into this category. No legal obligations apply, but voluntary codes of conduct are recommended.

Examples

Spam filters in email programs

AI-powered product recommendations in online shops

AI optimization in computer games

Automatic spell checking

Obligations

No legal obligations. Voluntary codes of conduct and transparency towards users are recommended (Art. 95).

Quick Test: Which Risk Category?

QuestionIf yes...Risk category
Does your AI system manipulate human behavior?System is likely prohibitedProhibited AI
Does it affect health, safety or fundamental rights?System qualifies as high-riskHigh-Risk AI
Does it generate content or interact with users?Transparency obligations applyLimited Risk
None of the above apply?Voluntary measures recommendedMinimal Risk

Source: Regulation (EU) 2024/1689 of the European Parliament and Council of 13 June 2024. EUR-Lex Volltext

Find out which risk category your AI systems fall into.