This research aims to develop and design a hybrid model for predicting the bankruptcy of Iranian banks from 1390 (2011-2012) to 1402 (2023-2024). This innovative model combines artificial intelligence (AI), machine learning techniques, and a spatial econometrics approach to improve prediction accuracy and consider spatial dependencies and risk contagion among banks. Therefore, this study utilizes Bayesian Neural Networks, Gaussian Neural Networks, Cerebellar Model Articulation Controller (CMAC), Multilayer Perceptron (MLP), Reinforcement Gradient, and Ant Colony Optimization algorithms in the AI and machine learning sections, and a spatial autoregressive model in the econometrics domain. Initially, financial and macroeconomic data were collected and cleaned. Then, the aforementioned models were evaluated using accuracy, Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), coefficient of determination (R-squared), and AUC-ROC. Furthermore, the spatial relationship between banks was examined using the spatial autoregressive model to explain the role of spatial dependency in the spread of bankruptcy risk. The research results demonstrated that combining the spatial autoregressive model with Bayesian Neural Networks and Multilayer Perceptron yields remarkable performance in predicting bank bankruptcy. Also, variables such as cost-to-income ratio, capital adequacy, liquidity, loan-to-asset ratio, and bank size, as internal factors, play a significant role in estimating the probability of bankruptcy. On the other hand, macroeconomic variables, such as GDP growth, inflation rate, and economic cycles, have a considerable impact on exacerbating or mitigating bank risk. The spatial autoregressive model also confirmed that risk contagion among neighboring banks could intensify the risk of bankruptcy. Based on the findings of this research, integrating AI and machine learning methods with spatial models provides a comprehensive and effective approach for predicting the bankruptcy of Iranian banks. Liquidity management, cost control, monitoring high-risk loans, and attention to macroeconomic dynamics will be most effective when geographical considerations and risk contagion are also taken into account. Therefore, it is recommended that monetary policymakers and bank managers utilize these hybrid models in the bank monitoring process and improve the stability and health of the banking system by strengthening financial structures and implementing reform policies.