Fraud poses a significant threat to the sustainability of enterprises and organizations. Effective prevention of enterprise/organization fraud is crucial in organizational behavior (OB). Small and medium enterprises/organizations are particularly susceptible to sudden and unpredictable frauds perpetrated by certain key individuals. In contrast, larger enterprises benefit from more robust internal controls, especially those listed and traded over-the-counter (OTC), making them less prone to abrupt fraudulent activities. However, fraud within these large entities can severely impact shareholders, jeopardizing their investments. In the age of artificial intelligence, deep learning algorithms can detect and predict more effectively. This research employed deep neural networks (DNNs) and recurrent neural networks (RNNs) to develop a fraud detection model for listed and OTC companies. The research subjects were 114 Taiwan-listed and OTC companies from 2005 to 2022, including 38 companies with fraudulent activities and 76 without. Findings revealed that the DNN model outperformed the RNN model across various metrics, including accuracy, precision, sensitivity, specificity, F1 score, and AUC. Additionally, the DNN model exhibited lower rates of both type I and type II errors compared to its RNN counterpart. This research successfully developed two models suitable for fraud detection, contributing to academic research and practice related to fraud. The insights garnered from this research serve as a reference for academic researchers, CPAs, enterprise risk assessment and rating agencies, investment advisers, securities analysts, corporate management, and government research units.