← Back to stories Business person holding a scam alert sign over a laptop, warning against online fraud.
Photo by Gustavo Fring on Pexels
凤凰科技 2026-03-09

Please Forward This Article to That Boss of Yours Who Wants to Install OpenClaw

Founder flags security and complexity risks

Li Nan, founder of NuMiao Technology (怒喵科技), has urged ordinary users to steer clear of grassroots OpenClaw installation drives that promise “free installs,” “free training,” or outsized returns. OpenClaw requires complex local deployment and high technical skill; non‑technical participants typically must either pay someone to install it or spend large amounts of time following online tutorials. Who wants that hassle — and at what risk?

Reported theft risk and short-term outlook

Li warned that free events are particularly risky. It has been reported that, given OpenClaw’s current security posture, user passwords, QR‑code account data and other credentials could be stolen in minutes. He also reportedly predicted that many of the tedious setup steps could be simplified within three months, and that six months from now it may be unclear whether people will still be using OpenClaw.

Safer alternatives and practical advice

For people who genuinely want to learn about AI and agents, Li recommended buying access to commercial models such as Claude (Anthropic), Gemini Pro (Google) or Grok (xAI) to master how to prompt systems correctly. For hands‑on experimentation he suggested Kimi Agent — reportedly developed by a reputable firm and already in widespread use among global developers — as a less risky practical option.

Broader context

Local deployment is appealing in China because cloud access and international model availability are shaped by geopolitics, export controls and corporate policy. But those factors cut both ways: the scramble to DIY more powerful tools can expose non‑technical users to fraud and data theft. The comments were posted on Phoenix New Media’s Dafeng Hao platform and carry the platform’s standard user‑upload notice.

AISpace
View original source →