China bans OpenClaw from government computers and issues security guidelines amid adoption frenzy — nation scrambles to rein in popular AI agent

China's central government has warned state enterprises and agencies not to install OpenClaw on office computers this week, according to Bloomberg, as multiple government bodies moved to rein in the Austrian-developed AI agent following a surge in adoption across the country. The Ministry of Industry and Information Technology's National Vulnerability Database (NVDB) has also published security guidelines, and the People's Bank of China has added a separate warning on AI in the financial sector, the South China Morning Post reported.

OpenClaw, developed by Austrian programmer Peter Steinberger, is an autonomous AI agent that automates tasks including email management, calendar scheduling, and travel check-ins. Its adoption in China has been rapid enough to acquire a nickname — "raising lobsters," a reference to the app's mascot — and Tencent, Alibaba, Baidu, and MiniMax have all launched compatible tools. But there are widespread concerns around the fact that OpenClaw requires broad access to user files and can communicate externally, potentially exposing host machines to cyberattack or data leaks if OpenClaw isn’t used cautiously.

The NVDB advisory, developed alongside AI agent providers and cybersecurity firms, tells users to run only the official latest version, minimize internet exposure, grant minimum permissions, and guard against browser hijacking. Prohibited practices include using third-party mirror versions, enabling administrator accounts during deployment, installing skill packs that require passwords, and disabling log auditing. The NVDB specifically flagged connecting instant messaging apps to OpenClaw as a risk that could grant excessive read, write, and deletion permissions over files.

Article continues below

Meanwhile, the People's Bank of China called at its annual technology conference in Beijing on Wednesday for AI in the financial sector to be managed in a "proactive yet prudent, safe and orderly” manner. The China Academy of Information and Communications Technology said the day prior that it plans to begin trialing AI agent trustworthiness standards on the likes of OpenClaw starting late March.

Curiously, these restrictions sit alongside active policy support for the same technology, with Shenzhen’s Longgang district currently seeking public feedback on a draft policy offering subsidies of up to 2 million yuan ($289,000) for OpenClaw app developments.

"Chinese regulators typically respond with extraordinary speed to threats from emerging technologies, but the rate of adoption of OpenClaw and other agentic tools is still outpacing them," said Kendra Schaefer, partner and director of tech policy research at Trivium China, speaking to Bloomberg.

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.