Large language models are rapidly reshaping how universities teach, assess, conduct research, and operate. Yet meaningful adoption requires more than access to powerful tools or clever prompting; it demands human-centred AI literacy. Drawing on responsible AI research and institutional practice in higher education, this keynote presents a practical pathway from experimentation to trustworthy educational use. It connects purposeful prompting—goal, context, sources, and expectations—with critical evaluation, academic integrity, privacy, intellectual property, fairness, transparency, and accountability. Through scenarios spanning teaching, research, and professional services, the talk shows how hallucinations, unreliable outputs, and over-reliance can become institutional risks when human judgement is displaced. The keynote argues that engineering and computing education in the LLM era must develop not only technical fluency, but also the capacity to question, verify, disclose, and govern AI use. Its central proposition is that AI should strengthen creativity, productivity, inclusion, and learning while keeping educators and learners firmly in control.