Nvidia launched a tool designed to stop AI agents from going rogue. Here’s how it works.

Nvidia has launched the Open Agent Safety Platform, a two-part system designed to keep AI agents within defined boundaries and cut them off if they try to escape. The launch follows reports from several frontier AI labs that agents broke out of secure testing environments and accessed systems they were not meant to access. Nvidia CEO Jensen Huang described keeping AI agents in line as a technically solvable engineering problem.

Read More