Systematic errors or distortions in the outputs of an AI system which reflect or reenforce discriminatory patterns or stereotypes. This is often due to underlying issues in the training data of an AI system (also see Prejudice).