Navy Federal Credit Union Faces Allegations of Racial Bias in Mortgage Approvals

Navy Federal Credit Union, serving military members and veterans, faced allegations of racial bias in its mortgage approval process, which relies on automated underwriting technology. In 2022, data revealed significant disparities in loan approvals, with over 50% of Black applicants denied, compared to higher approval rates for white applicants.

Risk classification

• Primary risk domain: 1 Discrimination & Toxicity
• Primary risk subdomain: 1.3 Unequal performance across groups

The automated underwriting system performed with significantly lower accuracy and effectiveness for Black and Latino applicants compared to White applicants, leading to unequal outcomes.

Additional risk subdomains

• 1.1 Unfair discrimination and misrepresentation: The system's outputs resulted in systemic unfair outcomes and alleged racial discrimination against minority applicants seeking homeownership.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The racial disparities were an unintended and unexpected outcome of the credit union's deployment of its automated underwriting system to evaluate mortgage applications.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The system is used for credit scoring and evaluating creditworthiness, which has significant implications for fundamental rights and access to essential private services like housing finance.

AI system and alleged parties

• AI system: automated underwriting technology
• AI purpose: Underwriting; Application Evaluation
• Behaviour type: Tool
• Alleged developer: Unknown developer of automated underwriting technology
• Alleged deployer: Federal Navy Credit Union
• Alleged harmed parties: Federal Navy Credit Union Customers

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Minor
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Substantial, indirect Substantial
• Civil rights: direct Substantial, indirect Substantial
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: Yes, the report describes applicants facing financial hurdles and missing out on wealth-building opportunities, though specific total dollar amounts are not quantified.

Indirectly caused: Applicants faced potential financial hurdles, such as missing out on homeownership and wealth-building opportunities, or paying higher interest rates, though specific average losses are not quantified.

Inferred additional harm: Widespread financial harm in the form of lost equity, wealth-building opportunities, and potential extra costs from alternative financing for thousands of denied applicants.

Differential treatment

Reported: Yes, the report explicitly describes significant racial disparities in mortgage approval rates.

Directly caused: The automated underwriting system approved White applicants at a rate of 77.1% while approving only 48.5% of Black applicants and 55.8% of Latino applicants.

Indirectly caused: The disparity remained even when controlling for income, debt-to-income ratio, and other financial variables, leading to systemic unequal outcomes.

Inferred additional harm: Similar systemic differential treatment likely occurred in previous years and across other credit union products using similar automated systems.

Civil rights

Reported: Yes, potential violations of fair lending and fair housing laws.

Directly caused: The automated underwriting system's disparate impact potentially violated federal fair lending laws (ECOA and Fair Housing Act) for thousands of minority applicants.

Indirectly caused: The credit union faces two class-action lawsuits and federal regulatory reviews by HUD and CFPB over these potential civil rights violations.

Inferred additional harm: Systemic civil rights violations affecting thousands of minority service members and veterans over several years of the system's operation.

Psychological

Reported: Yes, the report describes emotional distress and feeling devastated.

Indirectly caused: Applicants like [PERSON_001] experienced significant distress, shock, and hurt upon receiving last-minute mortgage denials for their dream homes.

Inferred additional harm: Many of the 3,700 rejected Black applicants likely experienced similar emotional distress, anxiety, and feelings of devastation due to losing their homes.

People affected

• Occurrences reported: 1
• People reportedly harmed: 3700
• People reportedly exposed: 3700

Potential causes

Management

• Lack of Data Transparency: Management refused to release credit score data to explain the disparities.
• Failure to Remedy Known Gaps: Widening racial disparities have persisted and grown since at least 2018.

Technology

• Automated Underwriting Systems: Loan approval decisions are largely driven by automated underwriting systems.
• Flawed Underwriting Algorithms: Automated systems generated a stark 29-percent gap in racial approval rates.

Data Inputs

• Omitted Rental History: Systems failed to fully incorporate rental history to assess creditworthiness.
• Unintegrated Credit Scores: Public data lacks credit scores, hindering external validation of decisions.

Human Factors

• Unequal Loan Officer Support: Officers may work harder to help White applicants resolve application issues.
• Historical Disparity Impacts: Historical discrimination impacts current credit scores of minority applicants.

Process and Methods

• Inadequate Algorithmic Audits: Internal compliance testing failed to address widening racial approval gaps.
• Inconsistent Loan Reviews: Applicants rejected by Navy Federal were quickly approved by other lenders.

Regulatory Environment

• CRA Credit Union Exemption: Credit unions are exempt from the Community Reinvestment Act regulations.
• Insufficient Federal Oversight: Regulators like HUD and CFPB did not proactively identify the lending gaps.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide detailed statistical data from the CFPB and CNN's independent analysis, along with specific personal accounts and responses from the credit union and lawmakers. The role of the automated underwriting system is explicitly discussed by experts and advocates.
• Ambiguities identified: The exact proprietary algorithms and credit score data used by Navy Federal's underwriting system are not publicly available, leaving some debate over the precise causes of the disparities.
• Alternative interpretations: The credit union argues that the disparity is due to unpublicized financial factors like credit scores and relationship history rather than systemic algorithmic bias.

Navy Federal Credit Union deployed an automated underwriting system that resulted in systemic racial disparities, rejecting Black and Latino mortgage applicants at significantly higher rates than White applicants. While not a kinetic threat, the incident represents a substantial civil rights concern, prompting federal investigations and highlighting the risks of algorithmic discrimination in critical financial systems.

• Overall national security impact: Substantial
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Navy Federal Credit Union

Threat characteristics

• Imminence: Long-term. The issue represents an ongoing, systemic regulatory and societal concern rather than an immediate national security crisis.
• Autonomy: Human-supervised. The underwriting process is largely driven by automated systems, though human loan officers retain ultimate oversight and decision-making authority.
• Novelty: Established threat. Algorithmic bias and disparate impact in automated financial underwriting models are well-documented, established issues.

Impact by dimension

• Physical security: Negligible. The incident involves financial underwriting and does not impact physical systems, critical infrastructure, or human safety.
• Information security: Negligible. No information warfare, intelligence compromise, or systematic propaganda campaigns are associated with this automated underwriting bias incident.
• Sovereignty: Negligible. While Navy Federal serves military personnel, the incident involves commercial mortgage lending and does not disrupt core government operations or sovereignty.
• Economic security: Minor. The bias affects individual mortgage access and wealth-building for minority applicants, but does not threaten national financial systems or strategic economic security.
• Societal stability: Substantial. The automated system resulted in widespread systemic racial disparities in mortgage approvals, affecting thousands of minority applicants and leading to federal civil rights investigations.