The rapid diffusion of generative artificial intelligence (AI) has challenged the established mechanisms for safeguarding academic integrity in higher education. Conventional tools built around the detection of textual overlap are of limited use against texts that are formally original yet not written by the student. The article examines how the meaning of academic integrity is shifting under generative AI, analyses the limitations of technical detection tools (high false-positive rates and systematic bias against non-native English writers), and compares three models of institutional response: prohibition, permission with disclosure, and integration. It is argued that a sustainable answer lies not in detection but in redesigning assessment and embedding an institutional AI policy within the internal quality assurance system, thereby making it a subject of external review. A set of practical recommendations for universities of the Republic of Kazakhstan is proposed.

Аcademic integrity, generative artificial intelligence, quality assurance, accreditation, assessment design, academic writing, higher education.