Nvidia releases Open Agent Safety Platform to monitor and govern agentic AI

Nvidia has launched the Open Agent Safety Platform, a governance system integrating OpenShell software with DPU-based silicon to secure agentic AI from testing to deployment. This reference design provides full-stack control across hardware, compute, and robotics systems. While offering cybersecurity advantages, analysts suggest it may not address all agentic problems. The platform, currently in beta, requires specific Nvidia SDKs and Linux environments, making it more suitable for enterprise ecosystems involving partners like Cisco and Microsoft.