Chinese cybersecurity company 360 unveils new AI tools: Tulongfeng, which it claims is “China's version of Mythos”, and Yitianzhen, to automate cyber defense
Chinese cybersecurity firm 360 Security Technology (601360.SS) has developed what it calls a domestic answer to Anthropic's Mythos …
Context & Ripple Effects
360’s announcement follows related coverage that its AI-powered agent was used to uncover roughly 1,000 previously unknown vulnerabilities, positioning the company’s new releases as an expansion of an existing AI-security effort rather than an isolated model launch.
The surrounding coverage places Tulongfeng alongside Sakana AI’s Fugu as claimed alternatives to Anthropic models amid a U.S. export ban. That makes the launch relevant to the emergence of regionally developed cybersecurity models, though the reported performance claims remain claims.
First-order effects
- 360 gains two named products to present to security customers: Tulongfeng as a cybersecurity model and Yitianzhen as a cyber-defense automation tool.
- The company more explicitly ties its product positioning to Anthropic’s Mythos, inviting direct comparison on cybersecurity-model capability and deployment.
Second-order effects
- Chinese security buyers and vendors have a more visible domestic model option to evaluate for vulnerability research and defense automation, potentially shifting evaluation away from access to foreign frontier models.
- Competing AI-security developers, including those highlighted alongside 360 in related coverage, face greater pressure to substantiate model-performance and operational-security claims rather than rely on broad frontier-model comparisons.
Third-order effects
- If these tools prove useful in production, cybersecurity may increasingly be organized around specialized, locally available AI agents for both finding vulnerabilities and automating defensive work, not only general-purpose models.
- Export restrictions and security-model access constraints could reinforce separate regional ecosystems for high-end cyber AI; whether that occurs depends on independently demonstrated capability and customer adoption.
The trend: This is one data point in the localization of AI cybersecurity capabilities as vendors build specialized alternatives where access to leading foreign models is constrained.