China restricts OpenClaw AI agent deployment in government and state enterprises
Chinese authorities issued warnings restricting the use of OpenClaw, an open-source agentic AI framework, in government agencies and state enterprises over critical security vulnerabilities. The framework grants large-scale access to private data and can communicate externally, exposing systems to potential attacks.
What happened
Chinese authorities issued warnings to government agencies and state enterprises, including major banks, restricting deployment of OpenClaw agentic AI applications on desktop computers due to security concerns.
OpenClaw is an open-source agentic AI framework that allows large language models to control system functions including email, calendar management, web browsing, file access, and messaging platform connections. The CERT China identified critical security vulnerabilities in the framework.
The platform requires exceptionally broad access to private data and can communicate externally, potentially exposing computers to malicious intrusions and data theft.
Reported impact
The security concerns could influence institutional trust and confidence in deploying agentic AI systems in China's public and government sectors.
Index analysis
A February 2026 SecurityScorecard STRIKE team report identified 40,000 exposed OpenClaw instances accessible from the internet at publication, rising to 135,000 within hours. Approximately 45% were hosted on Alibaba Cloud, with 37% of instances located in China, indicating widespread deployment using unsecured templates.
Governance controls
Chinese authorities recommended strict measures including system isolation and disabling automatic updates to mitigate potential security risks.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.