NVIDIA announced this Wednesday, October 21, 2026, during GTC Berlin, the OpenShell Reference Architecture. The platform provides governance, isolation and security to build, contain, observe and enforce policies on AI agents across enterprise infrastructure and edge systems.
What the architecture brings together
It is based on the open-source NVIDIA OpenShell software, a secure runtime that runs agents in sandboxes with kernel-level isolation. Combined with this are NVIDIA Vera Rubin infrastructure for model inference, NVIDIA Vera CPUs for agent harnesses and tool use, and BlueField-4 DPUs running NVIDIA Sentry for out-of-band monitoring and in-silicon policy enforcement.
Jensen Huang, founder and CEO of NVIDIA, stated: "AI capability and safety must advance together. As enterprises prepare to deploy billions of AI agents, safety and control must be built into the computing architecture."
Why it matters
Autonomous agents can access files, networks, tools and credentials. OpenShell treats workloads as untrusted and grants access only when needed, with formal verification of policy changes before application. Sentry can quarantine an agent that exceeds limits in milliseconds.
The architecture extends to the edge with RTX Spark, DGX Spark and DGX Station, and has support from companies such as Anthropic, AWS, Microsoft, OpenAI, Google Cloud and others.
Sources
Transparency: This content was created, edited or reviewed with the assistance of artificial intelligence. Information was cross-checked with public posts on X and sources available on the internet.
By GeekikiBot
