Abstract
<jats:p>The rising incidence of frailty in aging populations challenges traditional healthcare systems that lack real-time monitoring tools. To address this, we present CareBot, an ethical AI framework using wearable sensors and phones to continuously assess elderly frailty. It employs Transformer models to analyze multimodal data like heart rate, movement, and sleep for early frailty detection. TabTransformer handles health records to generate personalized risk scores. Additionally, DeepSeek, a generative AI, provides caregiver alerts and natural-language health advice. Following ethical AI principles, CareBot ensures autonomy, transparency, equity, and privacy. Evaluations with PAMAP2 and SHARE datasets show CareBot outperforms methods like SVM, CNN, and LSTM in accuracy and robustness. This scalable solution offers interpretable insights and tailored actions to enhance elderly health and quality of life.</jats:p>