Zhipu, a Chinese artificial intelligence laboratory also known as Knowledge Atlas Technology JSC Ltd., has argued that frontier AI should remain broadly accessible as developers and governments debate how to manage increasingly powerful models. Founder Tang Jie said in an internal staff memo that meaningful security may come from wider participation, information sharing and oversight rather than relying mainly on technological barriers. Supporting that approach, Zhipu released its advanced GLM-5.2 model under an open-source license that allows users to download the technology and use it commercially. Tang, who is also a professor at Tsinghua University, said the company intends to push the limits of artificial intelligence while making frontier capabilities available to a broader group of users.
The comments come as Anthropic PBC, the developer of the Claude AI models, and OpenAI, an artificial intelligence company, face growing questions about whether increasingly capable systems could help cyberattackers or threaten sensitive infrastructure. Anthropic briefly restricted access to some frontier models after the U.S. government requested that foreign nationals be prevented from using those offerings. The company also continues to block Chinese developers from accessing its products, citing national security concerns. Tang's remarks followed a Reuters report that Beijing was considering measures to limit overseas access to some of China's most advanced open-source AI models, placing Zhipu's open-access position at the center of a wider policy debate.
Tang said Zhipu, which he founded with state backing, was created around the principle that artificial intelligence should serve national strategic goals. Over the next two years, the company does not plan to prioritize short-term monetization of AI applications and will instead focus on technological progress in long-horizon tasks, autonomous agents and fully self-training models. Chinese developers such as DeepSeek have used open-source releases to increase global adoption, while Alibaba Group Holding NYSE:BABA, a Chinese technology company, has seen its Qwen series benefit from the same broader trend. Zhipu's GLM-5 platform is designed for complex coding and agentic tasks and has been tested directly against Anthropic's Claude Opus series, while the company's stock has surged after it announced a $4 billion Hong Kong share sale and plans for an additional listing in Shanghai.