Beijing, June 24 — At the ISC.AI 2026 conference, 360 Security Technology founder Zhou Hongyi announced the AI tool Tulong Feng, designed for automatic software vulnerability discovery. According to Zhou, the development is a direct response to the American model Mythos 5 by Anthropic.

Tulong Feng is integrated into the Yitian Tulong package, which also includes the automated protection system Yitian Zhen. Additionally, 360 launched the "Basalt Shield" coalition, aimed at supporting companies in critical infrastructure and the import-substituting IT solutions sector. The tool's claimed performance is impressive: it has already discovered 3,432 vulnerabilities, of which 105 have been confirmed by Chinese regulators, and several have been included in the national database as high-criticality. Independent verification of this data has not yet been conducted.

Strategically significant is that Zhou emphasized the rejection of copying the American approach, which is based on "the most powerful model, the highest-performance computing, and advanced chips." Instead, 360 is betting on AI agents that combine language models, expert knowledge from security specialists, and extensive vulnerability databases. This is a pragmatic and, in my opinion, more effective path for the real-world cybersecurity sector.

The launch of Tulong Feng comes amid tightening AI regulation in the United States. Previously, the U.S. government demanded that Anthropic suspend access to Fable 5 and Mythos 5 for foreign nationals, leading to both models being disabled for all clients. According to some reports, the reason was White House concerns about possible access to Mythos by groups linked to China. In response to these restrictions, OpenAI launched a specialized model, GPT-5.5-Cyber, for vulnerability discovery and patching, and also opened limited access to GPT-5.6 Sol, Terra, and Luna for a narrow circle of trusted partners.

In the same month, Anthropic accused operators linked to Alibaba of distilling Claude. These events highlight the growing geopolitical tension in the fields of AI and cybersecurity, where every new development becomes not only a technological but also a political statement.

Expert opinion: The development of Tulong Feng is not just a technological response but a clear signal of the formation of an independent Chinese stack in the field of AI security. Betting on an agent-based approach rather than a race for computing power could prove to be a more sustainable strategy in the long term, especially under conditions of limited access to advanced chips. However, the key question remains the tool's actual effectiveness and its ability to compete with systems at the level of Mythos in practice, not just in claims.