Nvidia CEO Jensen Huang calls OpenClaw first truly trustworthy agent runtime framework

At the Build a Claw event in London on October 15, 2026, Nvidia CEO Jensen Huang described OpenClaw as one of the most important models so far, saying it could benefit multiple industries, particularly robotics.

Huang called OpenClaw “the first truly trustworthy agent runtime framework” and compared it to an exoskeleton-like large language model that provides a real-time runtime framework for LLMs. Motors, joints, wheels and robotic arms, he said, are just tools that ultimately respond to agents running on computers. He added that AI is reaching a practical turning point, with people using it for software programming, marketing agents, research assistance, lab assistance and drug discovery.

Huang also stressed the need for permission management and isolation for autonomous systems. Recalling early internet risks when browsers were not isolated from the underlying computer, he introduced Nvidia’s OpenShell system, which creates a containment layer within an operating system and follows a least-privilege principle. He said the company will keep monitoring autonomous systems, adding that alerts would be issued if a robot violates rules or acts beyond its permissions.