This paper proposes an AI- and IoT-based building safety management system that addresses the limitations of conventional fire alarm and surveillance solutions. Unlike existing methods—Where various sensors, CCTV, and door control operate independently—The proposed system integrates these components through a data relay device that collects and transmits real-time information on door status, sensor readings, and abnormal sound detections. An AI-driven abnormal sound detection algorithm analyzes sounds such as screams, explosions, or breaking glass, improving early warning capabilities even under conditions where visual surveillance may fail (e.g., due to smoke or darkness). Additionally, a cloud-based control platform enables remote monitoring and door control in real time, enhancing responsiveness in emergencies. In experimental evaluations, the system demonstrated high accuracy in detecting critical sounds (exceeding 92%) and maintained stable remote door control with a 98% success rate. It also efficiently processed data from multiple devices while keeping response times low, indicating strong potential for large-scale adoption. Despite these advances, challenges remain in refining the sound detection algorithm for noisy environments, strengthening cybersecurity against potential network threats, and validating performance across diverse building types. Nonetheless, this integrated approach significantly improves building safety by facilitating more reliable fire prevention, crime deterrence, and evacuation support representing a vital step toward future smart building and smart city infrastructures.