NVIDIA Opens Up AI Security with Open Agent Safety Platform

NVIDIA has introduced an open software platform called Open Agent Safety Platform to strengthen AI security from testing to deployment. The platform includes NVIDIA OpenShell, open-source software that provides a secure runtime boundary for controlling how autonomous AI agents execute tasks across open and closed models. This boundary is enforced by NVIDIA Vera CPUs, which are designed specifically for agentic AI.

NVIDIA Sentry, an out-of-band watchdog, runs on NVIDIA BlueField-4 DPUs to continuously monitor agent behavior. If an AI agent attempts to move outside its software boundary, Sentry quarantines and stops it in milliseconds. This continuous monitoring and enforcement provide a secure environment for agents to operate within.

The Open Agent Safety Platform is open-source, allowing organizations to extend it to work with third-party compute platforms, including those from Arm and Intel. Industry leaders are collaborating with NVIDIA to strengthen AI safety across various industries, including Anthropic, Cisco, CrowdStrike, Dell Technologies, Figure, HPE, Microsoft, Palantir, Palo Alto Networks, Perplexity, Red Hat, Salesforce, SAP, Scale AI, ServiceNow, and SpaceXAI.

The platform's goal is to accelerate discovery at the frontier of AI safety, ensuring that safety and security are addressed across the full agent stack. By providing open, customizable tools for enforcing more control over long-running agents, organizations can better manage the risks associated with autonomous AI systems.