AI Act Risk Classification

Definition

The four-tier risk framework established by the EU AI Act to regulate AI systems proportionately. The tiers are: (1) Unacceptable risk — AI practices that are prohibited outright, including social scoring and certain uses of real-time biometric identification (Article 5, applicable from 2 February 2025); (2) High risk — AI systems in safety-critical or fundamental-rights-impacting domains listed in Annex I and Annex III, subject to conformity assessment and extensive compliance obligations (Articles 6-49, applicable from 2 August 2026); (3) Limited risk — AI systems with transparency obligations, such as chatbots that must disclose their AI nature (Article 50, applicable from 2 August 2025); (4) Minimal risk — all other AI systems, which are unregulated but may follow voluntary codes of conduct.