OpenClaw, created in November by an Austrian coder, has gripped tech-savvy companies and individuals on the mainland, sparking "lobster fever" in reference to its red crustacean mascot.

It differs from bots like ChatGPT because it can execute real-life tasks such as sending emails, organising files or even booking flight tickets.

However, Chinese cybersecurity authorities and Beijing's ministry of industry and information technology have warned of the risks of OpenClaw hacks.

"We have noted that while Openclaw offers AI-assisted applications, it may also introduce certain risks, namely potential security risks like excessive permissions, data leaks, and system breaches," Hong Kong's Secretary for Innovation, Technology and Industry Sun Dong told reporters on Monday.

"In considering some of the uncertainties, especially the risks that OpenClaw brings, the Innovation, Technology and Industry Bureau has promptly reminded government bureaus at the present stage not to install OpenClaw apps on computers that are linked with the government network," he said.

After downloading OpenClaw, users connect it to existing artificial intelligence models of their choice and then give it simple instructions through instant messaging apps, as if to a friend or colleague.

The office of Hong Kong's Privacy Commissioner also warned on Monday that AI agents pose higher privacy risks than ordinary chatbots because they generally have more access to users' files and account credentials by default.

"If there are any vulnerabilities in the system design or safety control on these agentic AI with high-level access and access to multiple systems and data sources, it will pose significant risks to personal data privacy and data security as a whole," the privacy commissioner's office said in a statement.