NVIDIA Launches Open Agent Safety Platform to Secure Agents From Testing to Deployment
NVIDIA has released the Open Agent Safety Platform, a reference system design providing full-stack governance for AI agents. The platform integrates security controls across software, hardware, compute, and robotics layers.
Verified State Diff
Impact & Verification Analysis
AI developers, enterprise security architects, and robotics engineers deploying autonomous agents.
It establishes a standardized security baseline for agentic AI, reducing the risk of malicious exploitation in enterprise and industrial environments while accelerating the safe deployment of autonomous systems.
Full Fact Overview
The NVIDIA Open Agent Safety Platform functions as a comprehensive framework for securing autonomous AI agents throughout their lifecycle. By providing a reference architecture, it addresses the critical need for guardrails in agentic workflows, which are increasingly susceptible to prompt injection, unauthorized tool execution, and data exfiltration. This platform bridges the gap between high-level software security policies and low-level hardware execution, ensuring that governance is enforced at the compute and robotics infrastructure level rather than just the application layer.