Fair Lending & Adverse-Action Procedure (AI credit decisions)

Field	Detail
Organisation	[Organisation Name]
Document type	Policy — approved, mandatory
Framework basis	ECOA/Reg B §1002.9 · FCRA · Fair Housing Act · EU AI Act Annex III 5(b)
Owner	[Policy owner — e.g. AI Governance Lead]
Approved by	[Accountable executive / Board]
Version / status	v1.0 — DRAFT for adoption
Effective date	[Date]
Next review	[Date — at least annually]
Classification	Internal
Prepared with	autogovern.io — AI governance & risk templates

1. Purpose

Ensure that AI/ML used by [Organisation Name] in credit decisions is fair, explainable and compliant — with accurate adverse-action reason codes and ongoing fair-lending testing.

2. Scope

Applies to any AI/ML system used to evaluate creditworthiness, price credit, or make/materially influence a consequential credit decision, including third-party scores.

3. Adverse-action requirements (binding)

1. Specific principal reasons (ECOA/Reg B §1002.9): every adverse credit decision must state the specific principal reasons — accurate and specific — regardless of model complexity. "The model is too complex to explain" is not a defense.
2. FCRA §1681m: where a third-party consumer report or score is used, provide an adverse-action notice identifying the source and the consumer’s rights.
3. Timing & delivery per Reg B/FCRA.

4. Reason codes & explanations

For each model, maintain a mapping from model drivers to plain-language reason codes. Prefer methods that yield accurate, applicant-specific reasons (e.g. reason-code ranking or counterfactual explanations). Validate that delivered reasons genuinely reflect the decision.

Reason code	Plain-language reason	Model driver
[R01]	[e.g. limited credit history]	[feature]

5. Fair-lending / disparate-impact testing

Test outcomes across protected groups on representative data:

• Disparate impact (four-fifths rule): flag any group with selection rate < 80% of the top group.
• Investigate flagged disparities; pursue less-discriminatory alternatives; document.

Jurisdiction note: the CFPB removed the ECOA disparate-impact "effects test" (Reg B rule eff. 21 Jul 2026), but disparate impact remains actionable under the Fair Housing Act, DOJ enforcement, and state law, and the rule is contested. Retain disparate-impact testing and treat ECOA DI as jurisdiction-dependent, not eliminated.

6. Governance & monitoring

Name an accountable owner; add a fairness gate to the model release pipeline; monitor reason-code accuracy, override rates and group outcomes on a schedule; escalate material fair-lending findings.

Regulatory mapping

Instrument	Requirement supported
ECOA / Reg B §1002.9	Specific-reason adverse-action notices
FCRA §1681m / §1681e(b)	Adverse-action with source; accuracy of consumer-report scores
Fair Housing Act / DOJ / state law	Disparate-impact exposure (retain testing)
EU AI Act Annex III 5(b) + GDPR Art. 22	High-risk credit AI; solely-automated decisions
FCA Consumer Duty (UK)	Avoid foreseeable harm from credit AI

Appendix — Adverse-action notice (template)

We are unable to approve your application. The specific principal reason(s) were: [reason 1]; [reason 2]; …. This decision was based in part on information from [consumer-reporting agency, if any], who did not make the decision and cannot explain it; you have the right to a free copy of your report and to dispute its accuracy. You may also request a review of this decision. Contact: [route].

---

Document control & approval

Version	Date	Author	Approved by	Summary of change
1.0	[Date]	[Author]	[Approver]	Initial adoption

Sign-off

• Policy owner: __________________________ Signature: ______________ Date: __________
• Accountable executive: __________________________ Signature: ______________ Date: __________
• Next review due: __________

This template was generated by autogovern.io as a professional starting point. Review and adapt it to your organisation, systems, sector and legal advice before adoption. It is an educational aid, not legal advice.