A systematic skew in AI outputs that reflects prejudices or imbalances present in training data or design.