The innovation and technology chief on Monday warned of security concerns linked to OpenClaw, saying all government units have been told to refrain from installing the artificial intelligence agent.

Sun Dong, the secretary for innovation, technology and industry, also said officials are aware of the potential risks of using the AI agent, including excessive permissions, data leakage and system intrusion.

The administration had drawn up a comprehensive set of IT security policies and guidelines for government units, requiring them to conduct strict assessments before installing any applications within internal systems, the minister added.

"Given the uncertainties brought by OpenClaw, especially the security risks associated with it, the Digital Policy Office has reminded all bureaus and departments not to install OpenClaw on computers connected to government network systems," Sun said.

"We also suggest non-government organisations and individual users to take sufficient security measures when using OpenClaw."

The open-source AI agent software is designed to carry out tasks on users' computer systems on their behalf, such as clearing one's inbox, sending emails, managing calendars and checking in for flights.

OpenClaw has surged in popularity on the mainland, and adopting the technology has been dubbed "raising a lobster", in reference to the app's lobster logo.

OpenClaw was developed by Peter ⁠Steinberger, an Austrian, and uploaded to GitHub, a cloud-based platform that acts as a central hub for developers, last November.

Steinberger, who built OpenClaw to help organise his digital life, was hired last month by ChatGPT maker OpenAI. (Additional reporting by Reuters)


Edited by Tony Sabine