China’s Zhipu AI, also known as Z.ai, has introduced its open-weight GLM-5.2 model, drawing attention from researchers who say it can rival Mythos in some bug-detection and cybersecurity use cases. Although GLM still trails leading systems from Anthropic and OpenAI on broader general-purpose tasks, its performance suggests Chinese AI developers are narrowing the capability gap with their U.S. counterparts at a rapid pace.
That progress is likely to intensify concerns in Washington, where officials have sought to limit China’s access to high-end AI models such as Anthropic’s Mythos and Fable, along with the advanced chips needed to train and operate them. The Trump administration has treated Mythos and similar systems, particularly those capable of uncovering software vulnerabilities, as potential national security risks. OpenAI’s recent launch of GPT-5.6 has also sparked worries over possible misuse, prompting the company to restrict access.
GLM’s open-weight design means users can download and run the model themselves on commercially available hardware. That accessibility makes it highly adaptable and valuable for advanced users who want greater control, but it also raises the risk that malicious actors could deploy the technology with minimal oversight.
