NVIDIA has introduced an Open Agent Safety Platform designed to help organizations secure and govern AI agents throughout their lifecycle, from development and testing to production deployment.
The platform is built around OpenShell, an open-source secure runtime that allows organizations to establish boundaries and policies for AI agents while they operate. NVIDIA also introduced NVIDIA Sentry, a reference architecture designed to continuously monitor agent activity and respond when agents move beyond defined security boundaries.
Key Highlights
OpenShell: Provides a secure runtime for deploying AI agents with configurable policies and controls.
NVIDIA Sentry: Monitors agent activity and can isolate agents when policy violations are detected.
Hardware-based security: Sentry uses NVIDIA BlueField-4 DPUs to help enforce security controls.
Cross-platform approach: OpenShell is designed to support different computing environments.
Enterprise focus: The platform targets organizations deploying increasingly autonomous AI agents across business and technology workflows.
Industry collaboration: NVIDIA says more than 100 organizations are working with it on agent safety technologies.
Why It Matters
As AI agents gain the ability to access enterprise data, applications, and tools, organizations need mechanisms to control, monitor, and secure autonomous actions.
NVIDIA's new platform focuses on adding security controls across the AI agent stack rather than relying only on safeguards within the underlying AI model.