China's 360 Security Unveils AI Cyber Tools to Rival Anthropic's Mythos

Executive Briefing

• Launched two AI models—Tulongfeng and Yitianzhen—targeting automated vulnerability discovery and cyber defense
• Claims Tulongfeng identified 3,432 software flaws, with 105 confirmed by the Chinese government
• Acknowledges a 20-30% capability gap versus US AI models but argues China cannot afford to wait for parity
• Positions strategy as team-based 'professional attack-and-defense' over reliance on a single elite hacker or model
• Reflects China's broader push to develop sovereign AI cybersecurity tools independent of US technology