Nvidia launched the Open Agent security platform to help users control the extent to which AI agents can access and operate. Agents can write code, use tools, and handle complex tasks for hours or even days, requiring access to data and systems while setting clear limits on how they can be used. Nvidia OpenShell manages the task execution authority of the agent. BlueField-4 and DOCA add monitoring and security control independent of the scope of the agent's control at the infrastructure level, and the Vera CPU provides computing power support for the workload.

Nvidia launched the Open Agent security platform to help users control the extent to which AI agents can access and operate. Agents can write code, use tools, and handle complex tasks for hours or even days, requiring access to data and systems while setting clear limits on how they can be used. Nvidia OpenShell manages the task execution authority of the agent. BlueField-4 and DOCA add monitoring and security control independent of the scope of the agent's control at the infrastructure level, and the Vera CPU provides computing power support for the workload.