Abstract
More advanced technologies are now required to deal with financial fraud because of the more intelligent and responsive systems available in the U.S. banking sector. This research focuses on the impact of AI on fraud detection systems within commercial banks in the U.S. through a systematic literature review of articles. Implementing and adopting various AI technologies, such as neural networks, support vector machines, decision trees, and graph neural networks, have demonstrated remarkable achievements in identifying complex fraud, monitoring in real-time, and reducing false positive rates. Results show that the integration of AI technologies with bank operational processes has improved detection accuracy and operational efficiency, enhanced customer loyalty, and met regulatory standards. Furthermore, the study reveals gaps such as a lack of model transparency, ethical implications, and information bias. The application of explainable AI (XAI) and hybrid solutions promises to address these issues. Incorporating these technologies facilitates adaptation to changes in internet-based financial transactions. It also helps eliminate gaps associated with current AI systems that are not fully controllable, due to the nature of information related to fraud. These systems reduce previously identified gaps in modern banking fraud prevention, complementing AI-based technologies. Additionally, the results emphasize the importance of secure infrastructures based on advanced AI fraud detection techniques. The findings contribute to broadening conventional perceptions of banking, allowing for sensible adaptation to the rapid growth and shifts in the financial world and digital economy.