Alignment
Alignment describes tuning an AI system to match the intentions, values, and safety requirements of the humans deploying it — as opposed to a model that is capable but behaves unpredictably or undesirably. It is a central research field within AI safety.