中文版本：“AI养龙虾”走红：OpenClaw 本地智能体热潮与安全、政策并行
China’s recent “AI lobster” meme—online shorthand for running the open‑source OpenClaw agent locally—has pushed the project into mainstream discussion over the past two days. CCTV (China Central Television) described OpenClaw as a locally deployable agent with persistent memory and proactive task execution, and noted that cities such as Hefei, Shenzhen and Wuxi/Changshu have issued supportive policies. The surge in attention is now forcing a parallel debate about security, cost, and how quickly China’s agent ecosystem can scale.
CCTV said China’s National Vulnerability Database logged 82 OpenClaw vulnerabilities from Jan 1 to Mar 9, including 12 critical and 21 high‑severity issues, while MIIT (China’s Ministry of Industry and Information Technology) warned that default or improper configurations can create high‑risk exposures. Science and Technology Daily reported that some users face monthly token bills as high as RMB 30,000, and noted Shenzhen Futian’s “government lobster” pilot uses a human “guardian” to supervise deployments.
Capital‑market responses have followed. China Securities Journal, via Eastmoney, said listed firms such as ThunderSoft, Beixin Yuan, UCloud and Rockchip have disclosed OpenClaw adaptation or deployment progress on investor platforms, signaling a “cloud + device” push across the supply chain. Separately, New Beijing News reported Geely Holding chairman Li Shufu calling AI development “very crazy,” arguing that errors are part of the refinement process while launching a new cross‑era talent program to address skills gaps.
Xinhua’s Securities Times article cited the State Council’s “AI+” action plan targeting smart‑terminal/agent penetration above 70% by 2027 and 90% by 2030. The change is that local AI agents are no longer a niche experiment but a policy and enterprise‑planning priority; what happens next depends on safer default configurations, standardized compliance checks, and a clearer cost‑performance curve.
