Cheating or being cheated? Academic integrity and a changing cheating habitus in international AI governance discourse

The rapid adoption of generative artificial intelligence in higher education is transforming how academic work is produced and legitimized, raising questions about the governance of academic integrity. Current discussions tend to frame integrity as student misconduct, giving less attention to shifts in institutional expectations and educational practices associated with AI-assisted learning. This study examines how international AI policy discourse constructs academic integrity in higher education and allocates responsibility for ethical AI use among institutional actors. Drawing on Bourdieu's theory of practice and Laclau's notion of floating signifiers, it proposes the changing cheating habitus as an analytical concept for examining how dispositions towards authorship and legitimate academic participation might change with the conditions of academic work. The study applies Critical Discourse Analysis to four UNESCO and OECD publications issued between 2023 and 2026, comprising institutional guidance, policy outlooks and a UNESCO-commissioned anthology. Fairclough's three-dimensional framework is used to examine how academic integrity is constructed across the textual, discursive and social dimensions of policy discourse. The findings suggest that responsibility, transparency, trust, authenticity and responsible use act as floating signifiers. They provide normative coherence but remain operationally underdefined across institutional contexts. This ambiguity is embedded in contemporary AI governance discourse. Its effects arise through the relationship between policy language, institutional interpretation and established academic conventions. Institutions promote AI integration through narratives of innovation while continuing to regulate academic legitimacy through assumptions of independent authorship and visible intellectual labour. This creates a structural mismatch in which students are encouraged to engage in AI-supported academic work while remaining accountable to integrity frameworks that have not evolved accordingly. The analysis also suggests that enforcement mechanisms may operate as academic capital filters if differences in academic capital affect the visibility of AI-assisted work. The study defines cheating and being cheated as concurrent structural outcomes arising from governance frameworks that encourage AI use while retaining accountability mechanisms developed before generative AI. It presents academic integrity as structurally mediated and negotiated through discourse. The analysis informs educational leadership and AI governance, especially in rapidly changing contexts such as the United Arab Emirates and the wider Gulf region.