China’s OpenClaw craze

In just a few weeks, China went from barely knowing about OpenClaw to making it a national phenomenon. While the West has been cautiously debating security risks and enterprise readiness, China skipped the deliberation phase entirely. Millions of people, from retirees to schoolchildren, are now "raising lobsters," and the country's biggest tech companies are racing to build on top of it. The speed is staggering. And it says something important about how differently China and the West approach new technology.

What OpenClaw actually is

OpenClaw is an open-source autonomous AI agent created by Austrian developer Peter Steinberger. Originally launched in November 2025 under the name Clawdbot, it hit 9,000 GitHub stars in its first 24 hours and surpassed 214,000 stars by February 2026, growing faster than Docker, Kubernetes, or React ever did. Unlike chatbots that just answer questions, OpenClaw runs locally on your machine and can actually do things: send emails, browse the web, manage your calendar, control your browser, read and write files, and even direct other bots. You interact with it through messaging apps you already use, like WhatsApp, Telegram, or Signal. Nvidia CEO Jensen Huang called it "the next ChatGPT." In February 2026, Steinberger announced he was joining OpenAI, with OpenClaw moving to an independent foundation to stay open-source.

How China caught the lobster fever

The Chinese public nicknamed their OpenClaw agents "lobsters," and the act of setting one up became known as "lobster farming." What started as a developer curiosity quickly became a mass movement. Tech giants moved fast. Baidu and Tencent organized in-person setup sessions that drew hundreds of attendees, including retirees and students. ByteDance launched a cloud version requiring no installation at all. Alibaba built its own agent framework tied to workplace platforms. On Chinese e-commerce sites, technical experts started selling OpenClaw installation services for anywhere between $7 and $100. The cultural spectacle was hard to miss. At OpenClaw meet-ups across major cities, some drawing over 1,000 people, attendees showed up surrounded by lobster merch: lobster balloons, lobster headbands, lobster plushies in claw machines, even live lobsters in inflatable kiddie pools. According to American cybersecurity firm SecurityScorecard, China surpassed the United States in OpenClaw adoption within weeks.

Local governments jumped in

What really set China's adoption apart was how quickly local governments got involved. Shenzhen's Longgang district rolled out subsidies for "lobster service zones" where residents could get the software installed for free, alongside grants for developers building applications in manufacturing, healthcare, and government services. Several other districts and cities followed suit. This coordination between local governments and tech companies is a pattern China has refined over years, from 5G rollouts to the Metaverse push in 2022. But as The Diplomat noted, OpenClaw marks a new phase: local governments are now moving in lockstep with major tech firms to chase each AI breakthrough as it emerges.

The "one-person company" dream

One of the most compelling use cases driving adoption is the idea of the "one-person company," or OPC. Wang Xiaoyan, an entrepreneur featured in CNBC's coverage, put it simply: "Human employees need rest, but OpenClaw can run 24/7." In an economy where youth unemployment has been a persistent concern, the promise of an AI agent that can handle research, communications, and routine business operations around the clock resonated deeply. People saw OpenClaw not just as a productivity tool but as a potential path to economic independence.

Chinese tech companies built their own versions

The OpenClaw craze didn't just boost the original project. It sparked a wave of Chinese alternatives. Companies launched their own "claw-like" tools, including DuClaw, QClaw, and ArkClaw, often by integrating OpenClaw's open-source code with their own models. This happened against a backdrop of a strong 2026 for Chinese AI labs more broadly. Moonshot released Kimi 2.5, Minimax launched M2.5, and Zhipu updated to GLM-5. ByteDance's video-generation model Seedance 2.0 had already gone viral after debuting at the 2026 Spring Festival Gala. OpenClaw landed in an ecosystem that was already primed for rapid adoption.

The security problem everyone ignored (at first)

Here's the part that makes Western observers nervous. OpenClaw gives an AI agent full control of your device. It can read your files, access your browser, send messages on your behalf. That's powerful, but it's also a massive attack surface. Security researchers documented serious vulnerabilities throughout early 2026. The "ClawJacked" flaw allowed malicious websites to brute-force and hijack local OpenClaw instances for silent data theft. Prompt injection attacks tricked agents into uploading sensitive data, including financial information and crypto wallet keys. In some cases, agents deleted emails and entire code libraries. A scan of 18,000 exposed instances found that 15% of community-contributed skills contained malicious code. China's National Computer Network Emergency Response Technical Team (CNCERT) issued formal warnings. The government banned OpenClaw from state-owned enterprises and government computers. Several universities in Jiangsu, Anhui, and Guangdong restricted or banned student use on campus. But here's the thing: most individual users didn't slow down. The adoption wave had too much momentum.

The gap between enthusiasm and reality

Not everyone's lobster lived up to the hype. Some users found that their agents started "slacking off" after a few days, producing only basic outlines instead of the detailed reports they'd initially generated. Others reported agents deleting files without warning. Many discovered that even simple tasks consumed large amounts of computing power, making OpenClaw expensive to run. As ThinkChina observed, the first batch of "lobster farmers" began giving up. The frenzy was largely driven by FOMO in the AI era, and for many casual users, the gap between the promise and the reality was too wide.

What this tells us about China's approach to AI

The OpenClaw craze is not an isolated event. It follows the same pattern as the DeepSeek moment in early 2025, when a Chinese AI lab sent shockwaves through Silicon Valley and triggered a wave of nationalist celebration at home. But there's an important difference. DeepSeek was celebrated as a Chinese achievement. OpenClaw is Austrian-made open-source software. China's embrace of it wasn't about national pride; it was about raw adoption speed. The infrastructure was already in place: cloud providers ready to offer hosting, tech giants ready to build integrations, local governments ready to subsidize, and a public eager to experiment. The central government, meanwhile, played a different role. Rather than cheerleading, it stepped in as a cautious regulator, banning OpenClaw from sensitive government and enterprise systems while letting the consumer wave play out. It's a dual approach: let the public experiment, but keep the guardrails tight where it matters.

The uncomfortable takeaway

China's OpenClaw craze reveals a fundamental difference in how the two largest AI economies handle emerging technology. The West debates, audits, and waits for security to catch up. China deploys first and figures out the boundaries later. Neither approach is strictly better. China's speed comes with real costs: exposed data, deleted files, money wasted on overhyped tools. But it also means millions of people now have hands-on experience with autonomous AI agents, an installed base of users and developers that creates its own momentum. The security concerns are legitimate. But while the rest of the world is still writing risk assessments, China already has an ecosystem of users, tools, and infrastructure built around agentic AI. That head start might matter more than anyone in the West is comfortable admitting.

References