Abstract
The rapid adoption of generative artificial intelligence (AI) in higher education has outpaced institutions' ability to integrate these tools in ways that preserve pedagogical alignment, assessment validity, and academic integrity. While large language models (LLMs) offer scalable support for learning and feedback, institutional responses are often fragmented, and tool driven. This paper proposes a curriculum-centric framework for AI integration that aligns AI use explicitly with intended learning outcomes, cognitive demands, and assessment evidence. Grounded in constructive alignment and human-centred AI governance, the framework mediates AI use through outcome mapping, assessment integration, rubric-aligned feedback, and continuous monitoring. A sample postgraduate module illustrates how AI affordances can be systematically enabled or constrained by curriculum intent. The analysis suggests that outcome-driven constraints support pedagogically appropriate AI use, strengthen assessment validity through process-based evidence, and embed governance as a core element of learning design.