⏎ Words Summary from News
**China must develop its own equivalent to Anthropic’s Mythos model to counter the cybersecurity risks of the AI era, warns 360 Security Technology founder Zhou Hongyi.** He described the US technology, which autonomously identifies software vulnerabilities, as a “cyber nuclear weapon” that has accelerated vulnerability discovery a hundredfold while slashing costs. Zhou made the remarks at a Beijing cybersecurity conference, noting that the US Project Glasswing alliance has granted over 40 organizations access to Mythos Preview—but excluded China, leaving the country at a strategic disadvantage.</p><p class="summary-lead">**Zhou argued that without a domestic Mythos equivalent, China lacks “reciprocal strategic deterrence capability” in cyber warfare.** He likened the situation to nuclear-armed states deterring one another, stressing that such a “game-changing weapon” cannot be held solely by the US. His warning is the first from a prominent Chinese tech founder about the strategic risks of frontier AI models, following Mythos’s global attention for its ability to autonomously exploit vulnerabilities.</p><p class="summary-lead">**However, Zhou acknowledged that Chinese large language models still trail US counterparts by 20 to 30 percent in core performance.** He urged China to avoid simply copying the US route and instead focus on AI agent systems that combine existing foundation models with specialist security data and vulnerability knowledge bases. Zhou pointed to 360’s new Tu Long Feng platform as an example, claiming its agentic architecture can already match Mythos in identifying vulnerabilities in open-source software and advanced AI systems.</p><p class="summary-lead">**The release of Mythos initially boosted shares of Chinese cybersecurity firms, including 360, but those gains have since reversed.** Shares in 360 Security Technology have fallen more than 30 percent since early May, reflecting market skepticism about near-term capabilities. Zhou’s call for a homegrown Mythos underscores the growing urgency for China to close the AI security gap, even as US restrictions tighten access to frontier models.
Key Takeaways
- China lacks a strategic deterrent in AI-powered cyber warfare as the US controls Mythos, a vulnerability-finding AI model.
- Zhou Hongyi warns that China’s exclusion from the US Project Glasswing alliance puts it at a severe strategic disadvantage.
- Chinese AI models still trail US counterparts by 20-30%, making a direct copy of the US route unfeasible in the near term.
- Zhou advocates for AI agent systems over frontier models, citing 360’s Tu Long Feng platform as a potential domestic alternative.
Insights & Analysis
- The Mythos model represents a paradigm shift in cyber offense and defense, effectively democratizing advanced hacking capabilities—a trend that will force all nations to invest in sovereign AI security tools or risk systemic vulnerability.
- China’s strategic pivot to AI agent systems, rather than chasing frontier model parity, could create a unique defensive niche, but it also risks deepening reliance on US-designed foundation models if not paired with indigenous breakthroughs.