Chinese cybersecurity company Qihoo 360 Security Technology says it has built two AI tools to rival Anthropic's Mythos. The rhetoric sounds like it was pulled straight from the Cold War.

Qihoo 360 founder Zhou Hongyi showed off the tools at a conference in Beijing. "Tu Long Feng" automatically hunts for vulnerabilities, while "Yi Tian Zhen" automates cyber defense. Tu Long Feng has already flagged 3,432 vulnerabilities, Zhou said.

Zhou puts the gap between top Chinese models and the most capable Western ones at 20 to 30 percent. 360 tries to make up for that with an agent-based approach that pairs models with security expertise and automated tools. "China cannot wait until model capabilities have fully caught up before beginning vulnerability discovery. We cannot afford to wait," Zhou said.

Separately, Jie Tang, a Tsinghua University professor who founded Z.ai and recently shipped the well-received GLM-5.2 model, estimated that a Chinese "Mythos" class model would arrive before Q1 2027.

Zhou frames AI vulnerability hunting as a nuclear deterrence problem

Without independent tests and benchmarks like those OpenAI and Anthropic provide, there's no way to verify Zhou's claims. But the rhetoric itself is worth paying attention to.

In his speech, Zhou went straight for the nuclear comparison. He called Mythos's ability to find vulnerabilities and build attack chains on its own the equivalent of "cyber-nuclear weapons of the AI era."

"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. China needs an "equivalent strategic deterrent capability." A weapon "that can shift the entire balance of attack and defense must not be left solely in the hands of others."

Zhou warned against "one-sided transparency," too. The U.S. could use Mythos to scan Chinese systems for vulnerabilities while China stays blind. "With Mythos, it becomes a situation where the enemy is fast and we are slow, the enemy is many and we are few. You are still relying on a few security experts for analysis, while the other side has already replicated a group of hacker agents to work simultaneously."

He pointed to the export ban on Anthropic's Fable 5 as proof. Fable 5 is the "civilian, neutered version of Mythos," and if someone jailbroke it, the whole world could potentially tap into Mythos-level capabilities. "This is what the U.S. government finds most intolerable. It must ensure that it alone possesses this capability, forming an absolute monopoly over this strategic asset," Zhou said.

The U.S. government uses the same national security argument to justify its own export controls on chips and Mythos. Both sides have spent years trading accusations over cyberattacks on critical infrastructure. It's the Cold War playbook, just applied to AI. One side builds a capability; the other says it can't afford to fall behind.

Even the names of 360's tools carry a whiff of war romanticism, likely referencing a classic Chinese martial arts novel where legendary weapons grant their owners supremacy.

Note: The Chinese transcript was translated using AI, and the excerpts were verified for accuracy using three additional AI systems.

AI News Without the Hype – Curated by Humans

Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.

Subscribe now