An AI system used to screen loan applications for a community development financial institution (CDFI) rejects applications from applicants with thin credit files at a high rate. The CDFI's mission is specifically to serve underbanked populations who typically have thin credit files. What is the MOST appropriate AI risk management response?

B is correct because the CDFI's institutional mission is to serve underbanked populations — the exact population the model is systematically rejecting. A model that is misaligned with the institution's core mission represents both a mission risk and a business risk. The appropriate response is to explore alternative data sources that better predict creditworthiness for thin-file applicants without simply relying on traditional credit bureau data. Accepting the rejection pattern (A) abandons the institution's mission. Increasing the cutoff (C) would worsen the problem. Eliminating AI entirely (D) is disproportionate; the issue is with this specific model's features, not AI generally.