# NVIDIA Sentry **Entity class:** Out-of-band AI-agent monitoring and enforcement reference design **Developer:** [[wiki/NVIDIA|NVIDIA]] **As of:** 2026-09-28 ## Definition **NVIDIA Sentry** is the hardware-isolated monitoring and enforcement layer in the [[wiki/NVIDIA Open Agent Safety Platform|NVIDIA Open Agent Safety Platform]] reference system design. Sentry runs on [[wiki/BlueField-4|NVIDIA BlueField-4]] data processing units and uses [[wiki/NVIDIA DOCA|NVIDIA DOCA]] to inspect agent requests and responses, maintain attested telemetry, verify agent identity and delegated authority, and enforce granular zero-trust access policies for data, tools, APIs and services. ## Enforcement position In NVIDIA Vera Rubin POD systems, BlueField-4 sits on the node's path to the model. That position gives Sentry an observation and intervention point outside both the agent process and host software. NVIDIA states that Sentry continuously monitors behavior, detects attempts to cross software boundaries and can quarantine or stop an agent in milliseconds. The isolation claim is architectural rather than psychological: Sentry constrains access and consequences without establishing what an agent understands or intends. Sentry is optional. [[wiki/NVIDIA OpenShell|OpenShell]] can provide runtime governance without BlueField-4, while Sentry adds a separate in-silicon trust domain that remains operational even when the host or workload cannot be trusted. ## Relationships - **reference-design component of:** [[wiki/NVIDIA Open Agent Safety Platform|NVIDIA Open Agent Safety Platform]]. - **paired runtime:** [[wiki/NVIDIA OpenShell|NVIDIA OpenShell]]. - **runs on:** [[wiki/BlueField-4|NVIDIA BlueField-4]]. - **built with:** [[wiki/NVIDIA DOCA|NVIDIA DOCA]]. - **domain:** [[wiki/Cybersecurity|Cybersecurity]], [[wiki/AI Safety|AI Safety]], and [[wiki/Security Engineering|Security Engineering]]. ## Related Articles - [[articles/Safety Discourse Is Transition Discourse|Safety Discourse Is Transition Discourse]] — places this node within the managed transition from concentrated machine advantage to governed public agent deployment. ## Sources / Provenance - [NVIDIA — Open Agent Safety Platform](https://www.nvidia.com/en-us/solutions/ai/agent-safety/). - [NVIDIA Newsroom — platform launch, September 28, 2026](https://nvidianews.nvidia.com/news/open-agent-safety-platform). - [NVIDIA Technical Blog — continuous in-silicon agent monitoring](https://developer.nvidia.com/blog/nvidia-open-agent-safety-platform-a-reference-for-continuous-in-silicon-agent-monitoring/).