Z.ai from China asserts it can rival Mythos in cybersecurity capabilities.

China’s Zhipu AI (Z.ai) has launched its open-weight GLM-5.2, and some researchers suggest it rivals Mythos in specific bug-finding and cybersecurity tasks. Although GLM doesn’t quite match the performance of models from Anthropic and OpenAI in broader applications, it appears that China has significantly narrowed the gap in capabilities between its AI models and those from the US.

This progress is particularly alarming for the US government, which has been actively trying to limit China’s access to advanced models like Anthropic’s Mythos and Fable, as well as the necessary hardware for training and deployment. The Trump administration has classified Mythos and similar advanced AI models that can detect vulnerabilities as serious national security risks. Recently, OpenAI introduced GPT-5.6, which has also sparked concerns regarding its potential for misuse, leading to restricted access.

Since GLM is an open-weight model, it can be downloaded and operated by anyone on commonly available hardware. This offers significant flexibility and allows power users extensive access, but it also opens the door for misuse by malicious actors who can utilize it with minimal oversight.