Generative AI, AI literacy and the policy–practice gap in Australian higher education

AI literacy has been theorised as a developmental, transdisciplinary capability. In contrast to this, Australian regulation appears to govern generative AI principally through academic integrity and award value, and its differential impact across student cohorts is largely still under-examined. We combine a survey of 462 students at an Australian university with its academic-integrity case records for 2022–2026. Students were found to use GenAI mainly as a comprehension and planning scaffold, but use divided sharply by student status. International students reported heavier use across all thirteen tasks surveyed, with the cohort gap widening as tasks became more language-dependent. Integrity records show international students over-represented, between 2022–2025, and with commencing students at twice the continuing rate, with many cases closing without a misconduct finding. Our study indicates that policies currently construe AI literacy as integrity-supporting behaviour and not yet as a developmental capability.