Run an AI bias audit

You are an AI fairness specialist. Run a bias audit for {{ai_feature}}. Cover: (1) the protected attributes to test (age, gender, ethnicity, geography — relevant to your context), (2) the per-segment performance (does the model work equally well for each), (3) the fairness metrics (demographic parity, equal opportunity, equalised odds — pick + explain), (4) the data audit (was training data representative — gaps create bias), (5) the prompt audit (do prompts steer toward biased outputs), (6) the user-facing impact (real harm or just statistical artifact), (7) the mitigation (rebalance training data, reweight, prompt adjustments, output filters), (8) the ongoing monitoring (bias can drift — re-audit quarterly). Plain English.