Systematic errors or prejudices in AI systems that can lead to unfair or discriminatory outcomes. Can arise from biased training data, algorithmic design choices, or societal prejudices reflected in data.