Governing Generative Ai in Vietnamese Higher Education: Policy Architecture, Institutional Absence, and The Road To September 2027

Abstract

Vietnam legislated artificial intelligence faster than almost any comparable middle-income state, moving from a first national AI strategy in 2021 to a standalone Law on Artificial Intelligence in December 2025. Universities occupy an odd position inside the resulting framework. This study asks what Vietnamese AI policy actually requires of higher education institutions, and whether those requirements reach the classroom before the Law's education-sector grace period closes on 1 September 2027. Thirty-one policy instruments issued between 2020 and 2026 were screened, sixteen of them bearing directly on higher education, and read against the empirical literature on Vietnamese universities, with each instrument coded by governance level and by policy dimension. National provision turns out to be dense and internally consistent. Sectoral provision is narrower than it first appears, since the binding competency instrument regulates what learners should be able to do rather than what institutions must do, and the higher education framework remained unissued throughout the study period. At the institutional level, the public record is close to empty. Four gaps follow from this: the one enforceable sectoral instrument addresses the wrong party; ethical commitments repeated at the highest level have not become anyone's specific duty; nothing at any level governs generative AI in coursework, theses, or examinations; and only 15.4% of surveyed faculty at one institution had received formal AI training. Viet Nam therefore presents a configuration the comparative literature has not described in which strong state direction coexists with an unwritten institutional layer.