Abstract: The rapid advancement of AI large model technology is bringing about an intelligent transformation in network and data security. This paper explores the innovative practices of AI large models in network and data security, outlining five evolutionary stages for AI implementation in cybersecurity. Antiy Labs' "Hengnao" security-specific large model significantly boosts performance by deeply processing core data, including vulnerabilities, malicious samples, and alert logs. The article also highlights the importance of building a secure and reliable intelligent agent development platform. This platform ensures the safe and efficient operation of intelligent agents through technologies like code sandboxing, file security detection, and multi-tenant isolation. In practice, intelligent agents have demonstrated immense value in accelerating alert analysis, threat hunting, threat intelligence, and compliance processes. They've achieved groundbreaking success, particularly in large-scale event security and data security.