Nvidia’s New Secure AI Tool, NemoClaw

Nvidia has rolled out a fresh tool called NemoClaw that promises tighter safety for its AI system, OpenClaw. Instead of running OpenClaw directly, NemoClaw puts it inside a protected “sandbox” that keeps the data and commands separate from the rest of the computer. This setup helps prevent accidental leaks or misuse while still letting the AI do its work. The system uses Nvidia’s Agent Toolkit to make the transition easy: a single command installs everything needed, including OpenShell. OpenShell supplies open‑source models and the sandbox environment that enforces privacy rules and network restrictions.

By adding this extra layer, Nvidia aims to give autonomous agents the power they need without compromising security. The move comes after growing concerns about how AI systems handle sensitive information and operate on public networks. Developers can now test NemoClaw in their own projects, and Nvidia plans to keep updating the toolkit with new safeguards as AI technology evolves.

Actions