Skip to content
← All compliance guides·Industry Guides
🏦Critical Risk

AI Compliance for Financial Services

Decades of financial regulation stacked on top of new AI law

Financial AI operates at the intersection of new AI-specific law and decades of existing financial regulation. Credit scoring AI must comply with ECOA, FCRA, and CFPB guidance before it even reaches EU AI Act requirements. DORA adds operational resilience requirements. Each regulatory layer adds obligations.

Applicable regulations

EU AI Act — High-Risk Credit (§5(b)) and Insurance (§5(c)) AI

Critical Risk

Scope: AI for creditworthiness evaluation (Annex III §5(b)) and life/health insurance risk assessment and pricing (Annex III §5(c)) in the EU

Full conformity assessment, technical documentation, explainability mechanisms, bias testing, EU AI database registration

Deadline: December 2, 2027 (deferred from August 2026)

Equal Credit Opportunity Act (ECOA)

Critical Risk

Scope: All credit AI in the United States

Disparate impact testing across protected classes, adverse action notices with specific AI factors, documented business necessity for model variables

Deadline: Ongoing

Fair Credit Reporting Act (FCRA)

Critical Risk

Scope: AI using consumer reports for credit decisions in the US

Permissible purpose, adverse action notices identifying specific AI factors (CFPB: "model said so" is not sufficient), dispute resolution process

Deadline: Ongoing

DORA — Digital Operational Resilience Act

High Risk

Scope: Financial entities in the EU

AI included in ICT risk management framework, third-party AI vendor contracts with audit rights, incident reporting for AI failures

Deadline: January 2025 (in force)

State Insurance AI Regulations (Colorado, California, NY)

High Risk

Scope: Insurance AI models in regulated states

Colorado: unfair discrimination monitoring, model availability for review. Varies by state.

Deadline: Varies by state

What to do first

1

ECOA/FCRA compliance is non-negotiable — implement adverse action reason generation before anything else

2

Test every credit model for disparate impact across race, gender, age, national origin

3

DORA ICT risk management: document every AI system as an ICT risk, update vendor contracts

4

EU credit/insurance AI: start conformity assessment work now — the Annex III deadline (December 2, 2027) allows less time than it looks like once you account for multi-system assessments

5

CFPB-compliant adverse action: identify top contributing factors, not generic explanations

6

Model drift monitoring: recertify or retrain when performance degrades

Estimated compliance cost

$50,000–$150,000 initial + $15,000–$40,000/year ongoing

Proactive compliance typically costs 3–5× less than post-enforcement remediation.

Generate your financial services AI compliance plan

Aegis Firma maps your specific AI systems against all applicable regulations for financial services — and generates prioritized documentation across 169 jurisdictions.

Get Financial Services compliance plan

Further reading

Other industry guides