Safety
NVIDIA Open Agent Safety Platform: A Reference for Continuous In-Silicon Agent Monitoring
The NVIDIA Open Agent Safety Platform blends the open‑source OpenShell runtime—executing autonomous AI agents in kernel‑isolated sandboxes—with hardware‑based enforcement via NVIDIA Sentry on BlueFiel
The NVIDIA Open Agent Safety Platform blends the open‑source OpenShell runtime—executing autonomous AI agents in kernel‑isolated sandboxes—with hardware‑based enforcement via NVIDIA Sentry on BlueField‑4 DPUs, creating a multi‑layered safety architecture. It enforces five core principles: verifiable policy, out‑of‑band enforcement, controlled access to the model, scalable authority guided by reasoning visibility, and shared responsibility across labs, enterprises, and hardware providers. The platform delivers continuous observability along the agent’s path to the model using DOCA correlation for real‑time policy enforcement at line speed.
Related
- Add Runtime Controls to AI Agents with NVIDIA OpenShell
- Operationalizing Reconstructive Authority: Runtime Construction, Dependency Resolution, and Execution Gating in Autonomous Agent Systems
Source: NVIDIA Developer | 2026-09-28