An AI audit team is reviewing a natural language processing model used to screen resumes. The team discovers the model assigns lower scores to resumes containing words statistically associated with women's colleges and female-dominated extracurricular activities. This is an example of which risk?

B is correct because the model has learned correlations between linguistic markers associated with women and lower selection scores, directly reflecting and amplifying historical discriminatory hiring patterns in the training data. This is a documented form of gender bias in AI hiring systems. Temporal bias (A) relates to data staleness, not encoded gender discrimination. Adversarial manipulation (C) involves intentional attack, not model learning. Distributional shift (D) describes performance degradation due to changed inputs, not the learned discrimination pattern.