Nvidia has launched its new open-source AI agent, NemoClaw, at GTC 2026, marking a shift toward agentic AI amid the global hype surrounding OpenClaw.
OpenClaw, an autonomous AI assistant capable of performing multiple tasks, has surged in popularity worldwide, including in China, where several companies are introducing their own variations.
During the keynote, Nvidia CEO Jensen Huang described OpenClaw as “the fastest-growing open-source project in history” and likened it to the operating system for personal AI, calling it the start of a “new renaissance in software.”
However, cybersecurity experts have raised concerns over OpenClaw’s security, warning that it can access private data, communicate externally, and be exposed to harmful content—a combination described as a “lethal trifecta.”
NemoClaw addresses these concerns by providing enhanced privacy and security features for AI agents. The platform enables organizations to deploy AI agents capable of automating complex, multi-step tasks for their workforce while maintaining strict control over data.
OpenShell: Enforces policy-based guardrails and gives users control over agent behavior and data handling.
Nemotron: Powered by OpenShell, this suite includes multiple AI models for tasks such as text generation and graph analysis.
Using a privacy router, NemoClaw agents can leverage cloud-based frontier models while integrating local computing resources, allowing the agents to learn new skills and complete tasks in line with privacy rules.
Deploy “self-evolving” agents across multiple environments with a single command.
Optimize local compute resources to run high-performance models locally, reducing costs and enhancing data privacy.
Currently available in an early preview for developers and organizations.
Nvidia positions NemoClaw as a more secure alternative to OpenClaw, giving enterprises a reliable way to harness AI agents while safeguarding sensitive data.
