The need for privacy in elderly care is crucial, especially where constant monitoring can intrude on personal dignity. This research introduces the development of a unique camera-based monitoring system designed to address the dual objectives of elderly care: privacy and safety. At its core, the system employs an AI-driven technique for real-time subject anonymization. Unlike traditional methods such as pixelization or blurring, our proposed approach effectively removes the subject under monitoring from the scene, replacing them with a two-dimensional avatar. This is achieved through the use of YOLOv8, which facilitates accurate real-time person detection and pose estimation. Furthermore, the proposed system incorporates a fall detection algorithm that utilizes a residual causal convolutional network together with motion features of persons to identify emergency situations and promptly notify caregivers in the event of a fall. The effectiveness of the system is evaluated to emphasize its advanced privacy protection technique and fall detection capabilities using several metrics. This evaluation demonstrates the system’s proficiency in real-world applications and its potential to enhance both safety and privacy in elderly care environments.