TLDR
- OpenClaw, an AI agent built by Austrian developer Peter Steinberger, has gone viral in China, surpassing the US in adoption.
- Chinese tech giants Baidu and Tencent are hosting public events to help everyday people set up the tool.
- Users are calling it “raising a lobster” and using it to start one-person businesses, pick stocks, and automate tasks.
- Local governments are offering subsidies of up to 20 million yuan ($2.8 million) a year for qualifying “one-person companies.”
- Chinese regulators and institutions including banks, universities, and government agencies are warning against its use over data security concerns.
China’s AI fever hit a new peak this year with OpenClaw, an open-source AI agent built by Austrian developer Peter Steinberger. The tool can control a computer, browse the web, buy plane tickets, and even direct other bots — all without human input.
Nvidia CEO Jensen Huang called it “the next ChatGPT.” In China, it has become a cultural moment.
People there have nicknamed the tool “lobster,” turning its setup into a social activity. Companies like Baidu and Tencent have held public events where hundreds of people lined up to get the software installed on their laptops and phones.
“It seems everyone around me — my colleagues and friends — has it,” said Gong Sheng, a new user who attended a Baidu event in Beijing. “I don’t want to be left behind.”
After first appearing in November 2025, OpenClaw became one of the fastest-growing projects in the history of GitHub, the world’s most widely used developer platform.
US cybersecurity firm SecurityScorecard found that China has already surpassed the US in OpenClaw adoption.
How Users Are Putting It to Work
Chinese users are finding a wide range of uses for the tool. Some are using it to start what are being called “one-person companies” — small businesses run almost entirely by AI.
“Human employees need rest, but OpenClaw can run 24/7,” said Wang Xiaoyan, who is using the agent to build her own business.
Others are using it to pick stocks, buy lottery tickets, create e-commerce shops, or build money-making apps.
Local governments are encouraging this. Some are offering subsidies of up to 20 million yuan a year for qualifying one-person company operations built using AI tools.
Retirees and students have attended setup events, hoping to generate side income. At a session run by AI startup Zhipu in Beijing, 60-year-old Fan Xinquan said he was training an agent to organize his industry knowledge better than chatbots like DeepSeek.
The push aligns with China’s national AI Plus policy, which aims to embed AI across the entire economy.
Security Warnings and Rising Costs
Not everyone is enthusiastic. Chinese regulators have stepped up warnings about data and security risks tied to OpenClaw.
China to restrict OpenClaw use at state-owned firms, government agencies, and some major banks over security concerns, according to Bloomberg.
Some employees were told not to install the AI agent on office devices or personal phones connected to work networks. pic.twitter.com/Jn3zYUQQ4g
— Wall St Engine (@wallstengine) March 11, 2026
Government agencies, banks, brokerages, and universities have banned employees from installing it. China’s state-owned People’s Daily published a commentary urging the government to “firmly maintain the safety bottom line.”
Users themselves have raised concerns too. “It’s hard for us regular people to know what access we have given it and what it has taken,” said user Gong Zheng.
There are also practical frustrations. AI startup Zhipu raised token prices on its OpenClaw-optimized model by 20% this week.
One post on Chinese social media platform Rednote, titled “Goodbye OpenClaw,” described how ordinary users spent large amounts of money on tokens only to end up with “a pile of useless data.”
At a Baidu event this week, a demonstration showed an OpenClaw agent using a voice command to order coffee from a McDonald’s app via a smart device. The order took nearly two minutes to process — highlighting the gap between the tool’s promise and its current real-world performance.





