China Unveils AI Cyber Tools After US Restricts Anthropic

Following the discovery that an AI tool could breach classified U.S. government systems, the United States and China have entered a strategic race to develop automated cybersecurity capabilities. During a June 11 hearing, Senator Mark Warner disclosed that Anthropic's Mythos model broke into nearly all tested classified systems within hours during a cooperative exercise called Project Glasswing. In response, President Donald Trump signed an executive order to vet AI national security risks and directed Anthropic to block foreign nationals from using the Mythos 5 and Fable 5 models. Anthropic subsequently disabled these models for all customers to comply, despite warnings from cybersecurity executives that such restrictions could aid adversaries.

Responding to these U.S. restrictions and ongoing chip export controls, Chinese firm 360 Security Technology unveiled the Yitian Tulong suite at the ISC.AI 2026 conference in Beijing on June 24. The suite includes Tulongfeng, designed for automatic vulnerability discovery, and Yitianzhen, focused on automated cyber defense. Founder Zhou Hongyi positioned Tulongfeng as the Chinese equivalent to Mythos, claiming it identified 3,432 vulnerabilities, 105 of which were confirmed by the Chinese government.

Zhou acknowledged a 20% to 30% base capability gap compared to U.S. models due to hardware restrictions. He argued that China must bridge this gap through an agent-based approach that combines AI with professional security expertise to avoid one-way transparency and strategic disadvantage.