Abstract: As artificial intelligence (AI) is integrated into physical education (PE), PE teachers are the primary users of these technologies. When applying AI, PE teachers face privacy risks arising from the leakage of physical activity data, fairness risks stemming from algorithmic decision-making intervening in PE teaching evaluation, and risks of intelligent decision-making undermining PE teachers' practical judgment, as well as risks of technological mediation weakening embodied interaction and affective education in PE classrooms. The boundaries of AI application should be defined from three main dimensions: technological application, teaching practice, and professional development. The study proposes that standardized strategies may be formulated in the future by improving relevant systems for PE teachers' application of AI, establishing a governance mechanism for their AI application, upholding a humanistic orientation in their use of AI, and comprehensively enhancing PE teachers' overall AI literacy.