Nvidia Unveils Open Agent Safety Platform to Secure Agentic AI
Nvidia has released its Open Agent Safety Platform, designed to monitor and govern agentic AI. The platform combines software with out-of-band DPU-based silicon in a reference system design that secures agents 'from testing to deployment.'
The Open Agent Safety Platform leverages Nvidia's OpenShell software, which provides a secure runtime boundary that traces all actions and enforces policy as agents run on Nvidia Vera CPUs. This approach aims to prevent the vast majority of agentic problems.
Nvidia's reference system design features NVIDIA Sentry, an out-of-band watchdog that runs on NVIDIA BlueField-4 DPUs to continuously monitor agent behavior. Sentry provides in-silicon security enforcement, quarantining and stopping misbehaving agents in milliseconds.