Banks can mitigate AI bias by understanding how data is collected and processed, testing training data for fairness, analyzing results for risk areas, engaging independent validators and using synthetic data. EY’s guidance notes that the first line of defence is understanding why data is collected and preparing it appropriately; it recommends measuring pre‑training bias using equality‑of‑opportunity and disparate‑impact tests, reviewing outputs with human experts, obtaining third‑party fairness audits and employing synthetic data to safeguard privacy and reduce bias.

See the related topic Why is AI explainability critical in financial services decisions?