Ensuring safety for workers is an important challenge in the industrial environment. This study introduced a computer-based approach to improve safety management and reduce workplace accidents on construction sites. The proposed system utilized the YOLOv11 algorithm to detect hazardous workers, particularly during steel structure installation at heights. The AI-based detector focuses on monitoring safety harness compliance, ensuring that workers adhere to safety regulations. The model was trained on a dataset of construction workers wearing safety harnesses, incorporating images from Vietnamese sites to capture variations in harness styles, shapes, colors, and working postures. The dataset was divided into 67% for training, 24% for validation, and 9% for testing, with YOLOv11 used for object detection. Experimental results demonstrate the system’s effectiveness in identifying dangerous positions, automatically detecting whether workers are wearing safety harnesses, reducing response time, and fostering a proactive safety culture. This study highlights the potential of real-time monitoring as a transformative tool for improving worker safety, ensuring compliance with safety standards, and enhancing safety risk management.