The use of Artificial Intelligence (AI) in civil engineering has seen rapid growth, yet its application to the reliability assessment of structural joints remains a relatively new frontier. This paper introduces a novel AI-based approach to evaluating the reliability of welded beam-to-column steel joints. With AI methods becoming increasingly integral to engineering workflows, this study demonstrates how they can be effectively utilized to assess the structural performance of joints, particularly by generating and analyzing moment-rotation relationships and identifying failure modes. A key aspect of the work is the development of a sophisticated, experimentally validated Finite Element Method (FEM) model that simulates realistic joint behavior. This model undergoes an extensive parametric study to expand the range of joint configurations, thus creating a robust dataset for analysis. Using this dataset, a physics-informed neural network (PINN) is employed and trained to predict the behavior of joints under various conditions. The PINN approach not only enables the generation of highly accurate moment-rotation curves but also facilitates the identification of failure modes across a wide range of joint characteristics. Once trained, the AI model is further used to augment data beyond the limits of the original training set, allowing for comprehensive evaluation of AI-driven predictions. The reliability of these results is then critically assessed to ensure they meet structural safety requirements. This work highlights the potential of AI-driven methods to offer significant computational time savings while maintaining high levels of accuracy and safety in structural design and analysis. The findings represent an important step toward the intelligent use of digital tools in structural engineering, suggesting that AI can substantially streamline workflows and improve the reliability of structural assessments in the future.