Using AI models to counter AI threats
As AI advances, the 360 Group has adopted a dual-track strategy of security and AI. It was among the first in the sector to propose using AI models to counter AI threats — building two lines of defense for large language model security and AI agent security.
Its independently developed AI security attack-and-defense system, Tulongfeng, became the first product in China to pass the China Academy of Information and Communications Technology (CAICT) assessment of AI agent capabilities in vulnerability discovery.
Tulongfeng has identified more than 15,000 vulnerabilities — including a Windows kernel vulnerability that had remained undiscovered for five years and a critical Office vulnerability that had gone undetected for eight years. The findings earned official acknowledgment from Microsoft.
The 360 Group has also helped international technology companies, including Google and Meta, to fix more than 200 vulnerabilities in AI frameworks, affecting more than 4 billion endpoint devices worldwide. More than 500 enterprises have now adopted Tulongfeng for specialized security testing.
For cyber defensive operations, the group has developed Yitianzhen — an automated security system that uses coordinated AI agents for threat monitoring, analysis, mitigation and response. It is helping to shift cyber defense from manual monitoring to autonomous operations, protecting digital infrastructure across industries.
