NVIDIA CEO Jensen Huang Praises OpenClaw and Stresses Strict Permission Controls
NVIDIA CEO Jensen Huang praised OpenClaw as a groundbreaking agent runtime framework for robotics during the Build a Claw event in London. He simultaneously emphasized the absolute necessity of strict permission controls and sandboxing mechanisms for autonomous systems. As artificial intelligence rapidly transitions into physical embodied systems and robotics, robust security and permission frameworks are essential to prevent autonomous agents from misusing hardware or going out of control. Huang's endorsement highlights OpenClaw's potential to bridge large language models with physical hardware safely. Huang compared the current AI agent ecosystem to the early internet era and introduced NVIDIA's OpenShell system, which uses kernel-level controls to isolate agents and enforce the principle of least privilege. For instance, an agent tasked with frying an egg only receives access to a frying pan rather than hazardous tools like a chainsaw.
## BACKGROUND
OpenClaw is an agent runtime framework that embeds a built-in agent loop, tool wiring, and prompt assembly to help large language models interact directly with software and hardware systems. NVIDIA OpenShell is an open-source security runtime designed to sandbox autonomous AI agents, using kernel controls to restrict file access, system calls, and network interactions.