In just a few weeks, OpenClaw, the AI agent that can independently navigate your device to send emails, monitor flight prices and find doctors appointments, has become a cultural phenomenon in China.
Tech analyst Rui Ma translated several personal vignettes about OpenClaw adoption from Jiemian, a Chinese business magazine: “That feeling of not being able to eat, not being able to sleep, tossing and turning all night, it had only ever happened to me before in romantic situations,” said one 27-year-old OpenClaw user in China.
This tension between obsession and the fear of losing control reflects the broader duality of the Chinese government’s response to the new technology. While local governments are jumping on the craze with plans for subsidized office spaces and startup grants, Chinese regulators have issued risk advisories and guidelines for preventing privacy breaches and financial losses.
Alfred Wu (吴木銮), an associate professor at the National University of Singapore, expects a ministerial-level regulation for AI agents like OpenClaw to be issued at some point. But for now, it remains unclear how exactly Beijing’s concern will translate into policy.
The introduction of technologies like OpenClaw and ChatGPT underscores how difficult it is for governments to adapt to the galloping pace of AI development. In China, AI governance at the local and national level reflects this tension, aiming for regulatory flexibility in these early stages in order to balance the competing priorities of innovation and security.
One way in which China achieves this flexibility is by creating a framework within which local governments can experiment. While the central government sets a strategic direction for AI development and defines its legal, ethical and political boundaries, local governments have room to create policy that encourages innovation and adoption, said Lu Xu (徐露), a senior lecturer at Lancaster University and Chinese law expert. Local experimentalism has been a feature of Chinese policymaking since the reform and opening-up era of the 1980s and 1990s, when China began integrating into the global economy. Localities serve as testbeds for policymaking so that the central government can make an informed choice about what gets implemented nationally and what does not.
Nowadays, local governments adopt the role of “middleman” to promote AI development, Xu said. For example, they might organize the bidding process for computational power between providers and users.
This model promotes competition among local governments to outperform their neighbors on tech innovation, said Xu. At the provincial level, Xu described the pressure that Jiangsu is feeling from Zhejiang because Zhejiang’s capital, Hangzhou, is becoming a leader in humanoid robotics. Unitree, one of the hundreds of robotics-related companies based in Hangzhou, made the robots that performed kung fu in China’s 2026 new year’s gala. Within Zhejiang province, Hangzhou’s neighbor, Suzhou, is also feeling the pressure: “We’ve always seen [Hangzhou] as peers but now they are so far ahead because of the so-called Six Little Dragons,” Xu said, referring to the six rocketing tech startups based in Hangzhou, including Unitree and Deepseek.
While central government regulators, like the Cybersecurity Administration of China, or CAC, tend to focus on mitigating the risks of emerging technology, there is also policy experimentation happening at this level. One example is the upcoming regulations on “humanized AI interactive services,” which are being drafted by the CAC. Xu pointed to Article 13 of the draft rules, which states that companies cannot provide AI services to elderly users that “simulate their relatives or specific relationships.” The rationale behind why elderly people should be deprived of this choice wasn’t provided, nor is it clear how regulators will define “elderly.” “The point is, China is not going to have that debate,” Xu said. “We’re just going to have the rule and if that rule causes negative consequences and too much controversy and it’s unpopular, it’s just a rule. We can change it.”
The central government is cautious about stifling innovation by implementing tech regulations too early, so policy tends to ramp up, further solidifying at higher levels of legal authority. Xu calls this a “bottom up” approach to policymaking. After the emergence of ChatGPT, for example, “China was the first country to react with the generative AI guidance regulations,” said Xu. The Interim Measures for the Management of Generative AI Services were issued by the Cybersecurity Administration of China at the ministerial level. These measures focus both on industry growth and content, requiring that AI outputs uphold China’s “core socialist values.” Further policy could move to the next highest rung of legal authority, which are regulations passed by the State Council. After that, a comprehensive AI law could be passed by the National People’s Congress.
Qiheng Chen, a non-resident fellow at the Asia Society, thinks that policy documents and relevant speeches from the past three years show that the central government has been placing a greater emphasis on innovation since the AI craze began. Regulators are also selective about enforcement, Chen said. China applies stricter regulatory scrutiny to the big AI firms, understanding that startups are more burdened by the costs of compliance. “The thinking is that if you can get big platforms to comply, it doesn’t have to worry about the small potatoes,” said Chen.
While there has been some discussion in China about passing a national level AI law, Chen doesn’t expect this to come into fruition anytime in the next few years. The central government first needs more clarity on what issues require regulation and how compensation should be addressed for people like copyright owners whose intellectual property helps train large language models, Chen said. Beijing also wants to see how tech competition with the U.S. pans out. “China doesn’t want anything like the EU AI act,” Chen said.
The approach is “more about giving enough space for innovation to happen and for adoption to happen. Once the technology settles into the next phase, regulation can follow.”
Leave a Reply