In short
- 360 founder Zhou Hongyi unveiled “Tulong Feng” at ISC.AI 2026, saying it is China’s answer to the Anthropic Mythos
- Z.ai’s GLM-5.2, which was released the week that Anthropic models went dark, outperformed Claude Code on the automatic benchmark at a risk of about $0.17 per acquisition.
- All of these moves come as Anthropic continues to negotiate with the Marketing Department on repurposing Mythos 5 and Fable 5.
The US banned Anthropic’s cybersecurity AI model Claude Mythos in the background external controls and a tested contract. China held a conference and announced that it had something like that – then another company uploaded an almost identical version of the internet for free.
Qihoo 360 founder Zhou Hongyi took part in ISC.AI 2026 in Beijing on June 24 and gave a specific message: “China’s cyber security companies must have their own Mythos.” He uncovered Tulong Feng – AI 360’s security consultant is calling on the Chinese brand-along with Yitian Zhen, an automated defense platform, and a new home security partnership called “Panshi Zhidun,” or Shield of Bedrock.
Zhou’s design pointed. Mythos is the “cyber nuclear weapon” of the AI age – an autonomous system that can find problems, analyze them, and build attack chains without human control. “US agencies can use Mythos to check your weaknesses, but you don’t have the right to check Mythos,” he said. Chinese companies are excluded GlassesAnthropic’s vetted partner program includes Microsoft, Apple, and other major technology giants.
According to Zhou, Tulong Feng found 3,432 threats, of which 105 were confirmed by China’s regulatory agencies and several were identified as high-risk by the national security database. Zhou said that the first aid method – combining unique models instead of betting on a single limit – eliminates any differences that still exist. “America has a Legend,” he told the audience. “China also has its own ‘Heaven-Sword Dragon-Saber’.”
Beijing based Z.ai created another controversy by releasing another. The lab, also known as Zhipu AI, abandoned the GLM-5.2 shortly after the US government drew Mythos 5 and Fable 5 do not open the Internet to foreign citizens. GLM-5.2 runs under the MIT license—no registration gates, no geographic restrictions, freely modifiable by anyone.
Cyber security numbers were popular. Semgrep’s analysis of direct knowledge analysis – a test to test if the model can detect unacceptable errors found in the code, which was created through the F1 metric that controls accuracy versus recall – put GLM-5.2 at 39%, ahead of Claude Code in the same test.
A separate review of Graphistry found it to be comparable to Claude Opus 4.8 in terms of flag drawing difficulty. Cost per acquisition: about $0.17, versus $1 for Claude’s move.
Z.ai co-founder Tang Jie called the Anthropic takedown “very sad.” The company’s technology director Qinkai Zheng was eloquent: “We want the brand to be accessible to everyone.” When Elon Musk predicted that China would not match Fable’s technology until Q1 2027, Tang replied: “It won’t take long.”
Daily Debrief A letter
Start each day with top stories right here, including originals, podcasts, videos and more.
