China Builds AI Cyber Weapons to Counter Anthropic’s Mythos
Chinese cybersecurity giant Qihoo 360 Security Technology has unveiled two AI tools it says are designed to go head to head with Anthropic’s Mythos, entering the debate around China AI cyber weapons with language borrowed straight from the Cold War.
Founder Zhou Hongyi introduced the tools at a conference in Beijing. The first, “Tu Long Feng,” autonomously hunts for software vulnerabilities and has already flagged 3,432 of them. The second, “Yi Tian Zhen,” automates cyber defense operations. Both are framed as China’s strategic response to what Zhou sees as a dangerous imbalance in AI offensive capability.
Zhou acknowledged that China’s best AI models still trail their Western counterparts by 20 to 30 percent. Rather than wait for that gap to close, Qihoo 360 is using an agent based approach that combines AI models with deep security expertise and automated tooling. “China cannot wait until model capabilities have fully caught up before beginning vulnerability discovery. We cannot afford to wait,” Zhou said in his speech.
The nuclear analogy was front and center. Zhou described Mythos’s ability to autonomously find vulnerabilities and construct attack chains as “cyber nuclear weapons of the AI era,” arguing that the only safe response is for China to build an equivalent deterrent. “Why has there never been a real nuclear war? Because both sides had nuclear weapons and deterred each other. The same is true in cybersecurity,” he said. A capability that can shift the entire balance of attack and defense, he warned, must not remain exclusively in foreign hands.
Zhou also raised the alarm over what he called one sided transparency, warning that Mythos could be used to silently scan Chinese systems for weaknesses while China lacks any comparable tool. He pointed to the export ban on Anthropic’s Fable 5 as evidence of how seriously Washington views this advantage, describing Fable 5 as the “civilian, neutered version of Mythos” and suggesting that even it represents a strategic asset the U.S. is determined to keep under its control.
The broader context around China AI cyber weapons mirrors the exact dynamic Zhou is describing. Both governments have spent years accusing each other of cyberattacks on critical infrastructure, and both use national security as justification for chip export controls and AI restrictions. Meanwhile, Tsinghua University professor and Z.ai founder Jie Tang has separately estimated that a Chinese Mythos class model could arrive before Q1 2027, adding another layer of urgency to the rivalry.
Independent verification of Qihoo 360’s claims remains impossible without third party benchmarks like those OpenAI and Anthropic publish. But the rhetoric coming out of Beijing signals that the AI security race is no longer just about model performance. It is increasingly being treated as a matter of national survival.
