In the context of AI, bias refers to systematic and unfair distortions in the output of AI models, often arising from unrepresentative or prejudiced data used in training. This bias can lead to skewed results or discriminatory practices, disproportionately affecting certain groups or individuals.