Жасанды интеллектті тәуекелге бағдарланған реттеудің конституциялық шектері: ЕО AI Act және Қазақстан Республикасының «Жасанды интеллект туралы» Заңын салыстырмалы талдау

Abstract

The article examines the constitutional limits of risk-based regulation of artificial intelligence from the comparative perspective of the EU AI Act and the Law of the Republic of Kazakhstan “On Artificial Intelligence”. The author proceeds from the premise that a risk-based approach is not, in itself, a guarantee of human rights: it merely allocates regulatory attention among different types of AI systems. A constitutional problem arises where the category of risk begins to replace the review of legality, necessity, proportionality and judicial contestability of interference with individual rights. The article shows that the EU AI Act combines the prohibition of unacceptable practices, the classification of high-risk systems, requirements for risk management, human oversight and the assessment of impacts on fundamental rights. The Kazakhstani Law “On Artificial Intelligence” also introduces a risk-based framework, including the classification of AI systems by degree of risk and level of autonomy, the duties of owners and holders to manage risks, and special mechanisms for trusted systems with a high degree of risk. At the same time, the Kazakhstani model requires constitutional fine-tuning: risk criteria must be public, reviewable and connected not only with technical safety, but also with human rights. The article substantiates that a minimum set of guarantees should apply regardless of the risk category assigned to a system where AI is used in public administration or affects a person’s legal status. The author proposes legal benchmarks for the further development of Kazakhstani regulation: assessment of the impact of AI on human rights, a public register of significant AI systems, human review, judicially reviewable risk classification and a prohibition on the automated reduction of the level of constitutional protection.