VibeusNVIDIA introduced its Open Agent Safety Platform on Sept. 28,... - Vibeus
NVIDIA introduced its Open Agent Safety Platform on Sept. 28, alongside more than 100 industry partners. The company describes it as an open trust layer for AI agents.
The stack combines OpenShell, an open-source secure runtime, with Sentry as an optional, independent hardware-backed layer. OpenShell runs each agent in a sandbox and turns operator instructions into verifiable policies. Teams can set boundaries for files, networks, tools, processes, and credentials; the runtime checks those limits before execution and enforces them while the agent works.
For organizations that want an additional control layer, Sentry extends monitoring and enforcement onto NVIDIA BlueField hardware. NVIDIA says its DOCA software connects that layer to OpenShell policies and correlates agent actions, policy decisions, and tool and data access into a contextual activity record. A DOCA gateway also checks each agent’s identity and delegated authority against its assigned scope.
The platform is optimized for NVIDIA Vera CPU and BlueField DPU systems, while NVIDIA says it is compatible with other hardware. On Vera Rubin POD systems, BlueField-4 sits on the node’s only path to the model, enabling out-of-band observability and real-time policy enforcement. NVIDIA says existing Vera and BlueField-4 systems can enable these protections with a software update.
