Nvidia Debuts 2-Layer AI Agent Safety Platform

Nvidia unveiled its Open Agent Safety Platform on September 28, 2026, a 2-layer system to secure AI agents from testing to deployment. The platform combines OpenShell, an open-source sandbox, with Sentry, a hardware-based monitoring layer on BlueField-4 DPUs, to enforce policies and intervene if agents deviate. This solution addresses growing concerns about AI agent security incidents in enterprise deployments.