China's OpenClaw obsession

THE LARGEST open-source project in GitHub history is not an operating system, a programming language, or a database. It is a lobster-themed AI agent built by an Austrian solo developer in about an hour. OpenClaw — originally published in November 2025 under the name Clawdbot — has in barely four months amassed over 264,000 stars on GitHub, surpassing Linux and triggering what may be the most consequential divergence in AI adoption between the world's two technology superpowers. While American enterprises debate security policies and compliance frameworks, China has turned OpenClaw into a national sport.

The numbers are staggering. Chinese usage of OpenClaw has already surpassed that of the United States, according to cybersecurity firm SecurityScorecard. The frenzy has added over $100 billion in market value to China's technology sector in a matter of days. MiniMax, one of the leading providers of tokens that power the agents, saw its market capitalization reach HK$382.6 billion — officially overtaking Baidu's HK$332.2 billion — despite generating roughly 239 times less revenue than the search giant. Tencent, Alibaba, Baidu, ByteDance, and Moonshot have all launched their own OpenClaw-compatible products: KimiClaw, AutoClaw, QClaw, MaxClaw, ArkClaw, and more. Local governments in Shenzhen and Wuxi have rolled out subsidies, free computing credits, and cash rewards to support OpenClaw-related ventures. Meetups in Hangzhou, Guangzhou, and Shenzhen have drawn hundreds of engineers, investors, students, and business owners apiece. Fans wear lobster hats. Elderly citizens queue at Tencent offices for installation help. They call it "raising a lobster."

Claws for concern

But beneath the plush crustaceans and stock-market euphoria lies a tension that should give both bulls and skeptics pause. OpenClaw is not a chatbot; it is an autonomous agent that operates a user's computer, manages their inbox, books flights, and executes shell commands — all with broad access to private data and the ability to communicate externally. Cybersecurity experts have described this combination of capabilities as a "lethal trifecta." Beijing has noticed. Chinese authorities moved this week to restrict state-run enterprises and government agencies from running OpenClaw on office computers, with employees at major banks instructed to report and potentially remove installations. The government is simultaneously promoting national AI development and scrambling to contain its security implications — a contradiction that neatly encapsulates the broader challenge of governing agentic systems.