NVIDIA Launches AI Agent Safety Platform
NVIDIA has introduced the NVIDIA Open Agent Safety Platform to provide full-stack governance and control for AI agents, their hardware, and robotic systems.
NVIDIA has launched the NVIDIA Open Agent Safety Platform, a new solution designed to enhance the security and control of AI agents, their associated hardware, and robotic systems. The platform includes NVIDIA's OpenShell security software, which sets boundaries for AI agents operating on CPUs and supports both open and closed-source models.
The OpenShell software is designed for low-resource utilization on the Vera platform, enabling AI agents to complete tasks efficiently while maintaining safety. As open-source software, OpenShell can integrate with third-party platforms, including those from Arm and Intel.
NVIDIA also announced NVIDIA Sentry, a watchdog system capable of continuously monitoring AI agent behavior. If an agent attempts to breach its operational boundaries, Sentry can isolate or halt its execution within milliseconds. Sentry operates on a BlueField-4 DPU, enforcing security policies independently at the hardware level.
Companies including Anthropic, SpaceX, and Scale AI are already collaborating with NVIDIA to implement OpenShell for their AI models and agents. This platform addresses the growing need for robust security and control measures in complex AI deployments.