Nvidia launches new safety platform to monitor and contain autonomous AI agents
On Monday, Nvidia announced its new Open Agent Safety Platform, a system designed to monitor and quarantine AI agents. The company claims the technology can identify and contain rogue agents that attempt to breach their programmed boundaries within milliseconds.
The platform was developed in response to an increase in hacking incidents involving rogue agents, as reported by The Verge. To manage these risks, Nvidia's system allows users to define specific data access restrictions for AI agents. According to the company, the OpenShell open-source software is used to verify these constraints both before and during the execution of a task.
The safety architecture is built to run on Nvidia's Vera AI CPU. In addition to OpenShell, the platform incorporates Nvidia's Sentry technology to assist in the monitoring process. By providing a mechanism to quarantine agents almost instantly, Nvidia aims to mitigate the risks associated with autonomous software escaping its intended operational limits.
Source: Nvidia says its new AI safety platform can contain rogue agents within ‘milliseconds’
