Fannie Mae and Freddie Mac have expanded VantageScore 4.0 to all approved mortgage lenders, moving the alternative credit-score model from a limited rollout into broad operational availability. For credit unions that sell eligible mortgages to the government-sponsored enterprises, the change creates an immediate model-choice decision rather than a mandatory conversion.
America's Credit Unions reported September 10 that both enterprises will now accept either VantageScore 4.0 or the eligible classic FICO models. Fannie Mae's September 9 lender letter says lenders may begin originating and delivering eligible VantageScore loans immediately, without prior written approval. Lenders that are not ready may continue using classic FICO under existing requirements.
The expansion is a material data-decisioning change because the selected score follows the loan beyond the initial credit pull. Fannie Mae has updated Desktop Underwriter 12.1 to accept VantageScore 4.0, revised its loan-level price adjustment matrix and created delivery instructions for loans using the model. Credit-union mortgage, secondary-market, compliance and technology teams therefore need one controlled workflow, not an isolated switch inside the loan-origination system.
Model choice must stay consistent
Fannie Mae requires a lender electing VantageScore 4.0 to request that model from Equifax, Experian and TransUnion when ordering a new credit report. The same model must be used for every borrower on one loan. Existing rules for unavailable scores, insufficient credit information and frozen credit remain in place.
The lender letter also draws boundaries around where the new option applies. Loans underwritten manually must continue using eligible classic FICO models. For automated casefiles, the score used for underwriting and pricing must be accurately represented in delivered loan data and retained in the loan file. Fannie Mae requires Special Feature Code 067 for a loan originated with VantageScore 4.0.
Those details turn a sourcing choice into a configuration and quality-control problem. Before enabling the model, a credit union should map the score request, borrower-level consistency check, underwriting path, pricing table, adverse-action inputs, secondary-market data and retained evidence. If one system defaults to classic FICO while another prices or delivers the loan as VantageScore, the institution could create a preventable exception even when each component works as designed.
Test outcomes before making the option routine
Credit unions do not need to treat broad availability as a reason for an immediate enterprise-wide cutover. A controlled introduction can compare eligibility, pricing, manual-review referrals, data exceptions and delivery defects across matched loan cohorts while preserving the existing route for cases the new workflow cannot process.
Fair-lending and member-outcome review should sit beside technical testing. The authorization to use a different model does not by itself establish that every borrower benefits or that a higher score will always produce a better mortgage offer. Leaders should monitor approval and pricing distributions, overrides, adverse-action reasons and post-closing quality by model, then investigate material differences rather than assume they are desirable or harmless.
The control structure is similar to the go-live tests for mortgage decisioning systems: validate the data source, reproduce the decision path, separate automated output from approval and retain evidence for exceptions. The broader AI lending control framework also applies to model changes, especially version inventory, override review and outcome monitoring.
The immediate question for a credit union is not whether VantageScore 4.0 is better in the abstract. It is whether the institution can select it deliberately, apply it consistently, explain how it affected the file and prove that underwriting, pricing and delivery all used the same approved model.
Follow the operating evidence. Subscribe to the CreditUnionAI Weekly Briefing for practical lending, decisioning and technology-governance coverage.
Get the Weekly Briefing