In 2025, a number of international organizations disclosed that: with the help of generative AI to automatically generate multi-language, strong simulation of the content of the conversation, the number of phishing attacks in some industries surged about 1200%, presenting the new characteristics of “large quantity, high quality, difficult to identify”. Attackers use stolen data to accurately construct the image of corporate executives and partners, bypassing simple keyword filtering and traditional blacklisting mechanisms. Meanwhile, AI-synthesized voice and video are used in phone fraud and deep forgery scenarios, causing the cost of manual verification to rise significantly. The security industry tries to identify suspicious patterns in advance at the conversation and attachment level through AI-driven email behavior analysis, content semantic detection, and account anomaly behavior modeling to realize “AI recognizes AI”. For enterprises, building a comprehensive “human-machine collaborative defense” that includes zero-trust access control, fine-grained authentication and continuous security awareness training has become a must for dealing with new-generation phishing attacks.