If an AI agent's operation is not on the permitted list or cannot be run in a sandbox, a separate classification sub-agent examines it. This sub-agent decides whether to allow the device call, try an alternative solution, or request approval from the user, increasing security and control.
Classification sub-agent decides on not permitted or non-sandboxable AI operations
If an AI agent's operation is not on the permitted list or cannot be run in a sandbox, a separate classification sub-agent examines it.
