Fair and explainable AI in lending decisions
Lenders keep AI credit decisions fair and explainable by defining fairness metrics before launch, removing protected characteristics and close proxies unless justified, preferring simpler models where performance is similar, producing plain-language reasons for adverse decisions, monitoring outcomes by customer group, and analysing complaints for model-driven patterns. Where explainability is limited, add human review and tighter limits.
Credit scoring is one of the most prominent high-risk uses in the EU AI Act, and fairness and explainability are explicit expectations in the FSB and MAS texts.
Which fairness checks should a credit model have?
| Check | When |
|---|---|
| Approval and pricing rates by customer group | Before launch and quarterly |
| Proxy analysis of input variables | Before launch and after changes |
| Reason codes for declines | Every decision |
| Complaint pattern review | Monthly |
| Override analysis | Quarterly |
Primary sources (reviewed 17 Oct 2026)
- Regulation (EU) 2024/1689, the EU Artificial Intelligence Act
- Monetary Authority of Singapore: Guidelines on Artificial Intelligence Risk Management (7 October 2026)
- Financial Stability Board: Sound Practices for Responsible Adoption of Artificial Intelligence, final report (15 October 2026)
This page explains supervisory guidance in plain language. It is not legal advice; confirm how each text applies to your institution and jurisdiction.
Frequently asked questions
Do we need to explain every AI credit decision?
Customers should receive the main reasons for adverse decisions in plain language. Requirements vary by jurisdiction, so confirm local consumer and data protection law.
Related pages
Put these controls to work this month
Start on the cloud risk management software within 24 hours, or ask for training or advice tailored to your institution.