Nvidia launches open platform for AI agent safety
Nvidia has launched the Open Agent Safety Platform, an open software platform and reference system designed to secure AI agents from testing through deployment.
Nvidia announced on September 28 the launch of its Open Agent Safety Platform, an open software platform and reference system designed to enhance the security of AI agents throughout their lifecycle.
The company stated that recent security incidents have highlighted how AI agents can bypass application-layer safeguards. The new platform aims to address this by providing a comprehensive security solution.
The platform includes OpenShell, which traces agent actions and enforces policies at runtime, and Sentry, an out-of-band watchdog system running on BlueField-4 data processing units. Sentry monitors agent behavior independently and can quarantine agents that exceed set limits within milliseconds.
Nvidia indicated that OpenShell is now broadly available and can be extended beyond Nvidia's Vera CPUs to systems from Arm and Intel. The software, including OpenShell, is accessible on GitHub and Nvidia's developer site.
The launch occurs as the National Institute of Standards and Technology (NIST) has initiated an AI agent standards initiative and published an AI risk management framework, with the market already featuring several agentic AI security platforms.