NVIDIA has introduced the Open Agent Safety Platform, an open security framework designed to keep AI agents within predefined operational boundaries.
NVIDIA has introduced the Open Agent Safety Platform, an open security framework designed to keep AI agents within predefined operational boundaries.
It combines OpenShell, which restricts access to files, services and data, with hardware-level monitoring designed to enforce controls while agents are running. NVIDIA says the platform is aimed at reducing risks such as unauthorized access, data exposure and agents acting beyond their assigned tasks.