An open-source autonomous AI agent software called OpenClaw has recently become very popular in China, with everyone from tech companies to ordinary consumers eager to try it out. This craze, referred to by users as "raising lobsters," is bringing to the forefront the policy tension in Beijing between applying artificial intelligence and maintaining data security.
Bloomberg and Reuters, citing informed sources, reported that the Chinese central government has advised state-owned enterprises and government agencies this week to avoid installing OpenClaw on office equipment. This is due to its extensive system access and autonomous execution capabilities, which may pose risks of data leakage, mishandling, and external attacks.
OpenClaw, developed by Austrian developer Peter Steinberger, is an AI agent tool that can autonomously complete tasks with minimal human intervention. It can perform tasks such as email sorting, schedule management, spreadsheet processing, and cross-application operations, which is a key reason for its rapid spread in China.
So far, China has not introduced specific written regulations for such agent-type AI, but regulation and standard setting are accelerating. The China Academy of Information and Communications Technology has announced that it will begin advancing relevant trustworthy testing and standard system construction from the end of March, indicating that the "spread first, regulate later" window is narrowing.
