China Orders State Banks and Agencies to Halt OpenClaw AI Over Security Risks

Chinese authorities moved to restrict the use of OpenClaw, an open-source autonomous AI agent, inside government agencies and state-owned enterprises including the country's largest banks, circulating notices that warned institutions against installing the software on office devices for security reasons.

Government agencies and state-owned enterprises received the notices in recent days, according to people familiar with the matter. Several institutions were instructed to notify superiors if the software had already been installed and to conduct security checks, with possible removal to follow. Certain employees at state-run banks and government agencies were banned from installing OpenClaw on office computers and on personal phones when those devices connected to company networks. One source said the restrictions were also extended to families of military personnel, though not all directives went that far. Some notices stopped short of an outright ban, requiring only prior approval before use.

The security concerns centered on OpenClaw's architecture as an agentic, or autonomous, AI system. Unlike conventional AI tools, OpenClaw requires unusually broad access to private data, can communicate externally, and could potentially expose computers to external attack, according to the notices. Those characteristics, which make agentic AI powerful for complex, multi-step tasks, also make it particularly sensitive in environments handling confidential government or financial data.

The Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission did not immediately reply to written questions.

The restrictions arrive as Chinese consumers and businesses have embraced agentic AI tools at a rapid pace. Over the past two months, OpenClaw attracted widespread experimentation across China, and several of the country's largest technology companies moved quickly to build products on top of it. Tencent, Alibaba, Minimax, Baidu and Zhipu all recently launched OpenClaw-based tools, positioning themselves for what many expected to be a fast-growing market.

Robert Lea, an analyst at Bloomberg Intelligence, said the regulatory scrutiny would likely constrain how unverified AI agents are deployed within state-owned enterprises and the broader government sector. He added, however, that the crackdown is unlikely to stall wider adoption across China. Lea also noted that the AI agent offerings from Chinese firms are currently loss-leading products, meaning companies are not yet profiting from the rush of interest.

The measures reflect a pattern in how Beijing has approached powerful new technology platforms, moving to contain security risks within sensitive state infrastructure while stopping short of policies that would broadly impede commercial development. By focusing restrictions on government agencies and state enterprises rather than the private sector, authorities appear to be drawing a line between national security environments and the wider consumer and business markets where OpenClaw has taken hold.

The variability in the notices, some mandating outright bans and others requiring only prior approval, suggests the directives were issued by multiple bodies or individual institutions rather than through a single centralized order, and that implementation will likely remain uneven across the sprawling network of Chinese state entities.