Nvidia announced on Monday, Sept 28, the NVIDIA Open Agent Safety Platform, designed to keep AI agents under control from testing through deployment, the company said. The platform combines open-source NVIDIA OpenShell secure-runtime software with the NVIDIA Sentry reference system design.

OpenShell creates a secure runtime boundary that traces every agent action and enforces policy on Nvidia's Vera CPUs, according to the announcement. As open-source software, it can be extended to third-party processors from Arm and Intel.

Sentry is an out-of-band hardware watchdog running on Nvidia's BlueField-4 DPUs, built on DOCA software. "Sentry provides in-silicon security enforcement, meaning that if an AI agent attempts to move outside its software boundary, Sentry quarantines and stops it in milliseconds," Nvidia said in a statement to SecurityWeek.

More than 100 organizations are working with the platform's technologies, Nvidia said, including Anthropic, Salesforce, SAP, CrowdStrike, Palo Alto Networks and Cisco. The platform is designed to enforce agent controls in silicon, so they apply even if the host machine is compromised.