If your AI model is surfacing a list of people who have access to a compromised system, how confident are you that the list ...