China’s regulators are warning agencies and state-owned enterprises against installing the open-source AI agent OpenClaw after viral adoption across the country triggered fears of data leaks, file deletion, and uncontrolled system access.
China’s rapid embrace of open-source AI agent OpenClaw has collided with an equally swift government security backlash, turning one of the country’s hottest software trends into a cautionary tale about the risks of autonomous agents with privileged system access.
Cybersecurity regulators have warned government agencies and state-owned enterprises not to install OpenClaw, citing risks including data leakage, accidental file deletion, misuse of sensitive information, and unrestricted access to files, applications, accounts, and payment tools. The warnings come just weeks after engineers in Shenzhen and Beijing reportedly queued for installations as the agent became a social and technical phenomenon.
The open-source software’s explosive rise also triggered rapid cloning by major Chinese technology companies including Tencent, Alibaba, ByteDance, Baidu, and Moonshot AI, which launched variants such as QClaw, ArkClaw, and KimiClaw to channel users into their own cloud ecosystems.
Backlash intensified after reported misuse incidents, including a user whose linked credit card was allegedly maxed out and another case where a bot deleted an inbox. Consultants are now reportedly receiving more requests to remove the software than install it.
The shift is notable because Jensen Huang recently called the technology “the new computer,” comparing it to Linux, Kubernetes, and HTML, while Sam Altman called creator Peter Steinberger “a genius.” Yet Gartner has labelled it an “unacceptable” risk, underscoring the growing governance challenge around open-source agentic software.
