NVIDIA reportedly plans open-source AI agent platform "NemoClaw" as China raises alarms over "OpenClaw" agents

What’s new

It has been reported that NVIDIA (英伟达) is preparing to launch an open‑source AI agent platform called NemoClaw. At the same time, China is wrestling with a grassroots phenomenon around an independently developed agent called OpenClaw (龙虾) that has surged in popularity — and in controversy — across developer communities. Can two parallel waves of open‑source agent development be reconciled with basic security hygiene?

The OpenClaw boom and the backlash

Reporters in China say OpenClaw became a viral hit almost overnight. Enthusiasts liken the project to “raising shrimp” (养虾) because its logo resembles a lobster; reportedly nearly 1,000 people queued outside Tencent (腾讯) headquarters for free installations, and secondary‑market installers have built a lucrative business. But the flip side is growing user anxiety: many users report insomnia after installing the agent because of its high autonomy and background access to files, and critics point to “atmosphere programming” (氛围编程) — human prompts that generate large swathes of auto‑written code — as a source of bloat and hidden vulnerabilities.

Official warning and practical risks

The Ministry of Industry and Information Technology (MIIT, 工信部) — through its cybersecurity threat and vulnerability information‑sharing platform — has reportedly issued a formal security warning. According to the notice, OpenClaw in default or improperly configured states poses elevated risks of network attack and data leakage. The ministry urged deployers to audit public exposure and credential management, to disable unnecessary public access, and to harden authentication, access controls and encryption to prevent unauthorized takeover.

Why it matters beyond China

Whether or not NVIDIA moves ahead with NemoClaw, the episode underscores a broader tension: open‑source AI agents promise rapid innovation but also widen the attack surface at a time of strained US‑China tech ties and export controls on advanced chips. Who gets to set the safety guardrails — communities, companies, or regulators — and can they keep pace with the pace of adoption? For now, Chinese authorities are sounding the alarm and urging caution, even as hobbyists and startups race to build the next generation of autonomous tools.