Help the user understand how often the AI system may make mistakes.
Allow the user to globally customize what the AI system monitors and how it behaves.
Inform the user when the AI system adds or updates its capabilities.