Secure online assessments

Strategies to mitigate generative AI risks in higher education

Abstract

The rapid rise of generative artificial intelligence (AI) tools has intensified universities’ long-standing challenge of ensuring that student work is authentic, an issue that is particularly acute in fully online programs. In 2024, new Tertiary Education Quality and Standards Agency (TEQSA) guidelines prompted our university to revise its Assessment for Learning Policy and Procedures, requiring every unit to have a secure assessment that is explicitly mapped to learning outcomes and validates core knowledge and skills. This study reports on the first pilot of that policy in three wholly online, postgraduate units. Using the 4-A Framework for secure assessment design, we implemented remotely-invigilated final exams in two units and a live, oral presentation in the third. Semi-structured interviews are planned with academics and students, and completion data and incident reports will be analysed. Preliminary pilot findings indicate that secure tasks can credibly validate student ownership of learning outcomes in an online environment. While both formats reduced academic-integrity breaches, oral presentations garnered higher perceptions of authenticity and lower than expected administrative overheads. We argue that secure, outcome-mapped assessments are feasible, scalable, and pedagogically valuable in online postgraduate education, and we outline policy and practice implications for institutions facing similar regulatory pressures.