Artificial Intelligence began as a bold idea in 1955, when John McCarthy and colleagues imagined machines that could learn, reason, and use language. Today, that vision is embedded in education through tutoring systems, analytics tools, automated grading, and virtual assistants. Yet most discussion still fixates on cheating. This keynote argues for a wider lens. The key issue is not misconduct, but what AI reveals about our assumptions around teaching and assessment. In other words: what have we really been measuring? Looking through major learning theories offers insight. Behaviourism highlights why AI excels at predictable tasks. Cognitivism is challenged by AI’s ability to generate knowledge. Constructivism positions AI as a support for meaning-making, while connectivism sees it as part of a broader knowledge network. Rather than a threat, AI can be a catalyst for change. It shifts focus from final products to learning processes, from standardisation to personalisation, and from compliance to curiosity and creativity. AI is not just a tool—it marks a pedagogical turning point. The challenge is designing learning environments where it enhances thinking, supports agency, and reshapes assessment into a meaningful reflection of growth.