NVIDIA Launches Open Agent Safety Platform to Secure Agents from Testing to Deployment
NVIDIA's new agent safety platform adds governance guardrails across testing, deployment, and hardware—closing a gap enterprise teams are scrambling to fill as autonomous workflows move into production.
Why it matters
NVIDIA is shipping an open reference platform for agent security and governance across the full stack—from testing through deployment on NVIDIA compute and robotics. This addresses a concrete operational gap for enterprises moving agents from pilots to production, where safety incident risk is rising.
The key facts
6 to knowPlatform announced Sept 28, 2026
Full-stack governance across software, hardware, compute, and robotics systems
Open software platform with reference system design
Targets agent testing-to-deployment lifecycle
Triggered by recent security incidents (not detailed in excerpt)
Integrates with NVIDIA compute and robotics hardware
Go to the source
EnterpriseAIhpcwire.com
Publisher excerpt: SANTA CLARA, Calif., Sept. 28, 2026 — NVIDIA today announced NVIDIA Open Agent Safety Platform, an open software platform and reference system design to strengthen AI security from agent testing to deployment, with full-stack governance and control across software and the hardware, compute and…