China Skips the Export List, Open-Sources the Receipt 🐉

Qihoo 360 founder Zhou Hongyi used the ISC.AI 2026 stage in Beijing on June 24 to declare that China's cybersecurity industry must build its own version of Anthropic's Claude Mythos, unveiling an AI vulnerability agent called Tulong Feng alongside an automated defense platform named Yitian Zhen and a new domestic security coalition, Panshi Zhidun ("Shield of Bedrock"). Zhou described Mythos as "cyber nuclear weapons" in the AI age—an autonomous system capable of finding vulnerabilities, analyzing them, and constructing attack chains without human direction—and noted that "U.S. organizations can use Mythos to scan your vulnerabilities, but you don't even have the right to look at Mythos," since Chinese companies are excluded from Anthropic's Glasswing vetted partner program, which includes Microsoft, Apple, and other major tech firms. According to Zhou's figures, Tulong Feng has identified a cumulative 3,432 vulnerabilities, with 105 confirmed by Chinese regulatory bodies and several flagged as high-severity by the national vulnerability database, with the agent-first approach of coordinating specialized models intended to offset any remaining base-model gap.

Beijing-based Z.ai, also known as Zhipu AI, chose a different route: releasing its GLM-5.2 model under an MIT license shortly after the U.S. government pulled Mythos 5 and Fable 5 offline for foreign nationals, with no subscription gates, no geographic restrictions, and free modifiability for any user. Performance benchmarks were notable. Semgrep's evaluation of insecure direct object reference detection, measured by the F1 metric that balances precision against recall, put GLM-5.2 at 39%, ahead of Claude Code in the same test, while a separate Graphistry evaluation found it matched Claude Opus 4.8 on a capture-the-flag challenge, and the company cited a cost per finding of around $0.17 versus over $1 for Claude-based workflows. Z.ai co-founder Tang Jie called Anthropic's withdrawal "deeply regrettable," while technical lead Qinkai Zheng stated plainly, "We want the model accessible to everyone." When Elon Musk predicted China would not match Fable-level capability until Q1 2027, Tang replied on X: "Won't take that long."

Meta separately introduced Brain2Qwerty v2 on Monday, an AI system designed to translate brain activity into text using non-invasive brain recordings, with the stated research goal of helping people who have lost the ability to communicate because of brain lesions.