NVIDIA put full-stack runtime safety controls around AI agents
Builders must treat AI agents as active systems requiring real-time runtime governance and cross-layer enforcement to prevent unsafe behavior, not just static model audits.
Frame 1 of 4
NVIDIA launches Open Agent Safety Platform combining software and hardware for AI agent governance
NVIDIA announced the Open Agent Safety Platform, which includes OpenShell open source software that traces agent actions and enforces policies on NVIDIA Vera CPUs, and Sentry hardware watchdog running on BlueField-4 DPUs that quarantines agents breaching boundaries in milliseconds. The platform aims to secure AI agents across software, compute, and robotics systems with participation from major industry players.