In recent years, artificial intelligence (AI) tools have been integrated into medical education at an unprecedented pace. The article reveals the dual nature of AI. On the one hand, AI can be used to support  education and develop clinical thinking skills in medical students. On the other hand, it risks violating academic integrity. It is shown that students of medical universities refer to large language data (for example,  ChatGPT) while preparing for classes, exams and writing papers. Key challenges such as AI hallucinations  (generation of plausible but incorrect outputs), algorithmic bias in AI systems and potential unfair use of AI for fraud and plagiarism are discussed. Modern approaches to detect AI-generated texts including integration of  special algorithms into similarity detection systems (for example, Turnitin) and Russian tools (AURA-Text system produced by Saint Petersburg State University) are presented. Using generative AI to simulate ‘virtual patients’  and clinical situations in education is given particular attention because it allows to safely practice diagnostic,  decision-making and communication skills. Examples of Russian (the Polymorbid Patient Platform of Pirogov  University) and foreign initiatives displaying the potential of adaptive and personalized AI-based learning are  provided. Strategic and ethical aspects of introducing AI into medical education, including the need to develop  guidelines, teaching students the principles of responsible use of AI and maintaining the balance between  technological innovations and following academic integrity are discussed.