The global competition for AI-driven cybersecurity supremacy has escalated. Chinese vendor Qihoo 360 has unveiled an AI system specialized in software vulnerability detection, claiming it is a superior alternative to Anthropic's Mythos model.

During the 14th Beijing Cybersecurity Conference, CEO Zhou Hongyi described Mythos as a "cyber nuclear weapon." He argued that U.S. restrictions preventing foreign nationals from accessing the model provide America with a strategic advantage to find flaws in software that other nations rely on, making an equivalent Chinese capability essential for deterrence.

Moving Beyond Brute Force

As detailed by The Register, Qihoo 360 is not simply trying to copy the Anthropic blueprint. Zhou criticized the "brute force" strategy—relying on massive models, extreme computing power, and top-tier chips—suggesting that China's approach focuses on efficiency over sheer scale.

Tulongfeng vs. Mythos

The Chinese system, named Tulongfeng, is designed to automate the discovery of software flaws. According to reports from Quartz, Tulongfeng has reportedly identified over 3,400 vulnerabilities, matching the performance levels attributed to Mythos.

This rivalry unfolds against a backdrop of high stakes. Anthropic's model, part of Project Glasswing, is regarded as the first AI capable of finding and exploiting vulnerabilities at scale, impacting major operating systems and browsers. While Anthropic restricts Mythos to a select group of partners due to safety concerns, China is aggressively pursuing similar tools to avoid being sidelined in the next generation of digital warfare.

Global Implications

The rise of autonomous AI bug-hunters is fundamentally altering the software security landscape. While these tools can accelerate patching, they also drastically reduce the time attackers need to exploit a flaw, shifting the balance of power toward near-instantaneous exploitation.