Named Entity Recognition (NER) is a cornerstone in natural language processing (NLP), aiming to identify, retrieve, and classify named entities into a predefined group. It is considered one of the most challenging sub-tasks, especially for the Arabic language, due to its complex nature and unique characteristics compared to English and other European languages. Machine learning (ML) and Deep Learning (DL) methods have been extensively utilized in NER research, leveraging the potential of deep neural networks and multi-task learning (MTL). Specifically, a MTL framework has been used to exploit high-motivated auxiliary tasks as part of speech (POS) to be included in a DL-based architecture. The proposed model is built based on convolutional neural networks (CNN), bidirectional long short-term memory (BiLSTM), and conditional random fields (CRF) to generate the final classification. The model can learn on both tasks that is NER and POS utilizing the KALIMAT dataset. The obtained results show that the proposed model outperformed the single-task model in performance, as the former of 97.91%. This result demonstrates the value of multi-task learning approach for recognizing named entities in Arabic language.