Jie Tang (탕제), co-founder of Chinese artificial intelligence company Z.ai, said cutting-edge AI should be made public as widely as possible and accessible to anyone. His stance runs counter to a development philosophy emerging as governments and major AI companies tighten access limits over the potential misuse of high-performance models.

On July 13, online media outlet Gigazine reported that Tang said in a recent internal memo, "One side pushes the limits toward higher intelligence, while the other must open the road to the future and broaden access to cutting-edge capabilities as widely as possible."

He set out a vision in which cutting-edge AI develops in a form that anyone can use, rather than being limited to specific companies or countries.

The remarks were made public shortly after reports said the Chinese government is considering measures to restrict overseas access to cutting-edge AI models. Debate in the AI industry has recently intensified over the security and openness of high-performance models. As AI models emerge that can find weaknesses in existing systems or be used for advanced cyberattacks, some countries are tightening access limits on national security grounds.

U.S. AI company Anthropic, after releasing 'Claude Mythos 5', a model that strengthened capabilities to respond to cyberattacks, restricted use by users outside the United States and by foreign nationals. The U.S. government controlled access, citing the model's potential to provide advanced cyber capabilities.

Criticism of such regulations has also continued. Technology experts in the United States and allied countries argued in an open letter that service blocks and access restrictions could instead create an environment that favors attackers over defenders. They said restrictions on services for Claude Fable and Mythos could work in favor of cyber attackers, and warned it could also dampen use for security research and defensive purposes.

Tang's comments also align with the relatively open development stance in China's AI industry. Z.ai is known as the developer of 'GLM-5.2', which has been reported to have recorded higher performance than Anthropic's Claude Code in a vulnerability detection benchmark. The company is known as a startup that began from Tang's lab at Tsinghua University, where he is a professor.

Chinese AI companies have been more active than U.S. competitors in pursuing open-source strategies. DeepSeek released its reasoning model 'R1' as open source, and Tencent has also recently announced a high-performance open model, adding momentum to expanding an open AI ecosystem.

Z.ai also presented a strategy of prioritising research and development over short-term profit. Tang said he would not make monetisation of the company's AI models the top priority over the next 2 years, and that it would focus on solving long-term tasks, autonomous agents and developing fully self-learning AI technology.

In the industry, the remarks are seen as more than a simple declaration of openness, and as an example of what technology strategy Chinese AI companies will choose amid moves to strengthen security and regulation. How far to allow the disclosure of cutting-edge AI, and what balance national regulation and corporate openness strategies will find, are expected to emerge as key issues for the global AI industry.