Government warnings follow rapid adoption
In recent days, Chinese government agencies and state-owned companies have begun telling their employees to avoid OpenClaw software. The warnings come after officials raised concerns about potential data security risks. According to people familiar with the matter, staff were asked not to install the software on work devices. In some cases, the guidelines also extended to personal phones and computers.
One source mentioned that employees of public companies had received instructions from regulators to avoid OpenClaw altogether. Another source from a government agency said there was no outright ban at their workplace, but staff had been warned of the safety risks and advised against the installation.
Security advisory highlights specific threats
The Technical Emergency Response Team of the National Computer Network/China Coordination Center has issued a security advisory regarding OpenClaw. They noted that improper installation and use have already led to several serious safety issues.
Among the main threats they highlighted was so-called “rapid injection.” This is where attackers embed hidden malicious instructions into web pages. If OpenClaw reads these pages, it may disclose sensitive information such as system keys. The advisory also warns of the risk of “misoperation,” in which OpenClaw could misunderstand user commands and accidentally delete critical data. This may include emails or important production information.
From early adoption to caution
OpenClaw was created by Peter Steinberger, an Austrian developer who published it on GitHub last November. OpenAI hired him last month. In China, the software achieved surprisingly rapid success. The phrase “raise a lobster” – a reference to the app’s lobster logo – has spread across Chinese social media. Large technology companies and some local governments quickly adopted the tool.
Tencent shares jumped 7.3% after the company unveiled compatible products. Startup MiniMax soared more than 20% as investors bet on the trend. Tencent launched Workbuddy, which connects to popular Chinese office applications. ByteDance introduced ArkClaw, a cloud-based version that requires no installation. Alibaba released CoPaw, which works with messaging platforms like DingTalk and Feishu. Zhipu AI has launched AutoClaw, making setup as easy as downloading a regular app.
Local governments were quick to follow as well. Shenzhen’s Longgang District has submitted a draft policy encouraging free deployment services and subsidies for developers. The Wuxi High-Tech District in Jiangsu Province announced subsidies of between 1 million and 5 million yuan for companies using the tool. This all happened as part of Beijing’s “AI plus” plan, which aims to introduce artificial intelligence to industries across the country.
Problems Arose Alongside Popularity
Rapid adoption has not been without its problems, however. A research center under the Shenzhen Municipal Health Commission held a training session last week that attracted thousands of participants. User complaints started coming in. The tool sometimes misinterpreted instructions, had weak access controls, and left people uncertain about where their data ended up.
It is still unclear how far the restrictions will go. There are questions about whether they will affect local grant programs related to OpenClaw. Shenzhen’s Futian district reportedly used the software to create an assistant for civil servants, according to state-run Southern Daily.
I think what’s interesting here is how quickly things have changed. One moment there’s this gold rush mentality around new AI tools, and the next there are security issues slowing everything down. It reminds me how new technologies often follow this pattern: rapid adoption, then caution, then perhaps more measured implementation.
Subsidies and government support were significant, so it is worth monitoring whether these programs continue. Maybe they will focus on different tools or modified versions that provide better security. The whole situation shows how complex deploying new AI tools can be, especially when dealing with sensitive information.
