The EU AI Act establishes a risk-based framework where high-risk AI systems (used in areas like healthcare, law enforcement, education, and critical infrastructure) are subject to mandatory requirements including security testing, risk assessment, and documentation.