AI security risks are evolving to be more multifaceted and intricate, with emerging threats presenting formidable challenges to the deployment of AI in enterprises. In response to the security concerns surrounding generative AI, corresponding solutions have been devised to establish a comprehensive security defense framework. This framework encompasses robust protection for both large-scale models and intelligent agents. Specifically, the security of large-scale models is fortified through a closed-loop mechanism that integrates "evaluation, defense, and operation," whereas the security of intelligent agents is bolstered by specialized products. These advanced security measures have found widespread adoption within the financial sector. Looking ahead, enterprises are expected to accord even greater importance to the holistic security of their AI application systems. Consequently, security evaluations will transition towards a model of continuous validation that spans the entire lifecycle of AI deployment.