The EU AI Act defines high-risk AI systems as those used in areas such as biometric identification, critical infrastructure, education, employment, law enforcement, and justice, and requires them to meet hard requirements including data governance, technical documentation, human oversight, accuracy, and robustness.