China's Qihoo 360 Launches AI Models to Compete with US Cybersecurity Tools

Qihoo 360 has introduced two AI models, Yitian Tulong, aimed at vulnerability discovery and automated defense, presented by CEO Zhou Hongyi at the ISC.AI 2026 conference. The Tulongfeng model identified 3,432 software flaws, with 105 confirmed by the Chinese government. Despite acknowledging a 20-30% capability gap compared to US models, Zhou emphasized the importance of a professional attack-and-defense team over a single advanced AI. The new models are designed to rival American tools, particularly Anthropic's Mythos, to enhance China's cybersecurity capabilities. The initiative reflects a strategic move for national deterrence amidst growing cybersecurity competition between China and the US.

Key Points: • Qihoo 360 unveiled AI models for vulnerability discovery and defense at ISC.AI 2026. • The Tulongfeng model identified over 3,400 software flaws, with some confirmed by the government. • The initiative aims to strengthen China's cybersecurity capabilities against US tools.