China’s government is telling its own officials to stay away from the tool its citizens can’t stop downloading. Asia Times reported today that Chinese authorities have warned government employees against sharing sensitive information with OpenClaw instances, framing the open-source agent’s file-access, web-browsing, and code-execution capabilities as a national security concern.
The warning lands in the middle of what NCT has been tracking as the fastest adoption-to-restriction cycle in recent tech history. OpenClaw went from WeChat phenomenon to state media alarm in under seven days. Now, the security apparatus is catching up to the enthusiasm.
What Asia Times Adds to the Picture
Previous reporting from Xinhua and Futurism tracked China’s tonal shift from enthusiasm to caution at the government messaging level. The Asia Times piece goes further, explicitly naming the cybersecurity risks that concern Beijing: OpenClaw agents can read local files, browse the web autonomously, execute arbitrary code, and — if configured carelessly — exfiltrate data to external servers.
For government networks, those capabilities represent exactly the kind of attack surface that China’s cybersecurity establishment has spent years trying to lock down. The irony is that China’s own tech companies — Baidu, Tencent, Alibaba — are simultaneously building commercial products on top of OpenClaw’s open-source framework.
No Comparable Regulatory Framework
Asia Times notes that China “has yet to introduce comparable rules” to the EU AI Act, which is heading toward an August 2, 2026 enforcement deadline for high-risk AI systems. China’s approach so far has been ad hoc: verbal warnings to government officials, state media signals about caution, and behind-the-scenes guidance to tech companies.
That gap matters. Without formal rules, enforcement relies on the kind of informal government guidance that can shift overnight. Companies building on OpenClaw in China have no regulatory clarity on what’s allowed, what’s restricted, and what triggers a crackdown.
The DeepSeek Parallel
The pattern echoes what happened with DeepSeek in late 2025, when Chinese authorities alternated between promoting the domestically-built model as a national champion and quietly restricting its deployment in sensitive contexts. The difference with OpenClaw: it’s US-origin open-source software, which makes the political calculus more complicated. Beijing can’t claim it as a domestic success story while simultaneously flagging it as a foreign security risk.
The tension is structural. China’s tech sector wants the capabilities. China’s security apparatus wants the control. OpenClaw, by design, makes that trade-off harder to manage than a closed-source API ever would.
Source: Asia Times, March 20, 2026