The Biased Algorithm

Your company uses an AI to screen job applicants. It has increased hiring efficiency by 90%. However, an audit reveals the AI is systematically down-ranking applicants from a specific minority demographic, not because of qualifications, but because it learned from biased historical hiring data. Disabling the AI would cause chaos and massive costs. Trying to fix it while it's live is complex and could take months, continuing the harm in the meantime. What is your immediate decision?