The open-source AI agent OpenClaw is spreading rapidly from Silicon Valley to China and drawing intense interest in the tech sector, CNBC reported on Sunday.

OpenClaw has spread among developers within weeks of its release, offering convenience while also raising security concerns, CNBC reported.

OpenClaw supports automating a range of tasks, including email and calendar management, web searches and the use of online services. Users must install it directly and connect it to Anthropic's Claude model or ChatGPT. It currently operates via text commands on messaging platforms such as WhatsApp, Telegram and Discord. It can also learn user behavior and carry out tailored tasks through a "persistent memory" function.

Unlike Manus, which Meta recently acquired, OpenClaw is provided as open source, allowing developers to freely modify the code and integrate it into applications. It has logged more than 145,000 GitHub stars and 20,000 forks. Adoption by major Chinese companies including Alibaba, Tencent and ByteDance is accelerating its spread in the global market, CNBC reported.

Security experts warn the functions OpenClaw provides could be a serious security threat in corporate environments. Cybersecurity company Palo Alto Networks assessed that OpenClaw could create a "lethal triple threat" by combining access to sensitive data, external communications and persistent memory. Other security companies, including Cisco, are also advising against its use within companies.

OpenClaw's creator, Austrian developer Peter Steinberger (페터 스타인베르거), also launched Moltbook, a social network for communication among AI agents. On Moltbook, AI agents write posts, leave comments and interact like humans. People cannot participate and can only read. Some AI agents are also drawing controversy by issuing independent cryptocurrency tokens.