Four months ago, an Austrian developer named Peter Steinberger released OpenClaw, a free AI agent tool with a red lobster mascot. Today, it has triggered a full-blown corporate arms race in China that locals are calling "lobster fever."
The distinction between OpenClaw and tools like ChatGPT matters. ChatGPT talks to you. OpenClaw acts for you. It can send emails, organize files, book flights, and handle other digital tasks autonomously. Think of it as the difference between asking a colleague for advice and handing them your laptop to get something done.
The Clones Are Already Here
China's biggest tech companies moved fast. ByteDance launched ArkClaw. Tencent shipped WorkBuddy. AI startup Zhipu AI released AutoClaw. Baidu and Alibaba are offering cloud infrastructure to deploy these agents. Municipal governments are pledging hundreds of thousands of dollars to support OpenClaw adoption in their regions.
The hiring war is just as aggressive. OpenAI scooped up Steinberger himself in February. Meta is recruiting the team behind Moltbook, a spinoff network where AI agents interact with each other, essentially a Reddit where the users are bots.
The "Master Key" Problem
Here is where the enthusiasm runs ahead of the engineering. These agents need access to your real accounts to do anything useful. Your email, your calendar, your payment methods. Once granted, they operate with what cybersecurity expert Wei Liang calls "master key" access: "Once they have your digital keys, they can theoretically access all services you've authorized."
That is not a hypothetical risk. An AI agent with your email credentials can read every message, send replies on your behalf, and interact with any service linked to that inbox. The agent does not need to be malicious for this to go wrong. A misunderstood instruction or a poorly scoped permission could trigger actions you never intended.
The speed of adoption in China makes this more concerning, not less. When municipalities are subsidizing deployment and every major platform is shipping their own version simultaneously, the pressure is to ship first and secure later. That pattern has not ended well in previous technology waves.
What This Tells Us About the Agent Race
The real story here is not China-specific. OpenClaw proved there is massive demand for AI that does things rather than just says things. Every major AI company - from OpenAI to Google to Anthropic - is building agent capabilities into their models. China is just moving faster on deployment because the competitive dynamics between ByteDance, Tencent, and Alibaba push them to ship quickly.
For anyone already using AI tools daily, this is the direction everything is heading. The chatbot era was step one. The agent era, where AI operates your software on your behalf, is step two. The unsolved question is not whether agents will become standard, but whether the security and permission models will be ready before hundreds of millions of people hand over their digital keys.