Artificial intelligence (AI) has become an essential part of daily human life with its various technologies. It can be used to generate creative ideas or to generate human-like texts. The challenge of recognizing AI-generated and human-generated texts is the basis of this research. Previous studies, which are very few, have proposed text detection models that only support the English language, which makes it difficult to handle Arabic and its linguistic complexities. Therefore, this research develops fine-tuning AI-generated text detection models that support both Arabic and English. The importance of this research revolves around the Arab and Saudi scientific community and contributing to the preservation of authentic human style and thought. Accordingly, this research followed the six-stage CRISP-MD methodology to develop the proposed model, conducting 33 experiments to achieve a highly accurate model. The experiments also included data extraction techniques such as Term Frequency - Inverse Document Frequency (TF-IDF) and Bag of Word (BoW). Accuracy was calculated before data balancing and after random undersampling. The BERT model was adopted for Arabic with 100% accuracy and the DistilBERT model for English with 98% accuracy.