Nvidia Unveils AI Agent Safety Platform With Hardware-Based Watchdog

Nvidia has announced the Open Agent Safety Platform, which the company describes as combining open source software with a reference system design. The announcement names two components: OpenShell, an open source runtime, and Sentry, a watchdog that runs on Nvidia's BlueField-4 data processing units.

OpenShell is described as a runtime that sandboxes agents and enforces policy on what they do. Sentry is described as a watchdog that runs separately from the agents it monitors, on Nvidia's BlueField-4 data processing units. Because Sentry runs on separate hardware rather than alongside the agent, the monitoring function is distinct from the environment in which the agent operates. The article does not state which policies OpenShell enforces, who authors them, or how they are revised.

The article does not report a release date, pricing, or any organisations that have adopted, tested or endorsed the platform. It does not state which agent frameworks or model providers the platform supports, nor whether it has been submitted to any standards body. The announcement describes coverage spanning testing through deployment, and pairs an open source software runtime with a hardware-based monitoring component.