Nvidia has introduced a system that allows AI agents — programmes capable of performing tasks independently — to be controlled. According to a statement from the company, the system makes it possible to define in advance which data an agent can access and what actions it can perform.

The system’s OpenShell tool monitors these boundaries and blocks unauthorised actions. Nvidia has also introduced an additional solution called Sentry, which can separately monitor an agent’s activity and isolate it if it attempts to operate outside the permitted boundaries.

OpenShell is already available, while Sentry has been presented as a design solution for an additional protection system.