Article

Nvidia launches AI an agent security platform that can isolate anomalous agents in real time

PANews September 28 reported that, according to the Securities Times, Nvidia launched the “Open Agent Security Platform,” which provides AI agents with two layers of security protection through two open-source software tools, OpenShell and Nvidia Sentry. OpenShell can run on Nvidia’s Vera central processing unit and restrict AI agents’ access permissions in real time; Nvidia Sentry runs on the BlueField data processing unit, where it can monitor agent behavior and isolate them when anomalies are detected, completing the process in as little as several milliseconds. Nvidia said the system is expected to help the AI industry test advanced AI systems more safely.

Source: https://www.panewslab.com/zh/articles/01a0e785-5cbc-717b-9419-1e50b68dff10