The CFPB Eliminated Federal Disparate Impact. Illinois Made It State Law. What Lenders with Illinois Customers Must Do Before January 2027.

TL;DR

• The CFPB eliminated disparate impact from federal ECOA enforcement on April 22, 2026, removing the effects-test language from Regulation B. Three months later, Illinois enacted SB 3777 (Public Act 104-0744), effective January 1, 2027, creating an independent state-law disparate impact standard for credit decisions under the Illinois Human Rights Act.
• A lender with Illinois borrowers can now face an Illinois Human Rights Act claim for a credit practice that produces discriminatory effects — even if the same practice faces no federal ECOA liability.
• Illinois didn’t define “financial institution,” so coverage likely extends to nonbank lenders, fintechs, and marketplace lenders with Illinois customers.
• Lenders using AI or algorithmic underwriting for Illinois borrowers need disparate impact testing documented before January 1, 2027. A practice that’s survived federal scrutiny post-April 2026 may still be actionable under Illinois law.

The CFPB moved in one direction. Illinois moved in exactly the other. The result is a state-federal divergence on one of the most consequential questions in consumer finance: whether a lending practice that’s facially neutral but has discriminatory effects violates the law.

That question now has two different answers depending on which regulator is asking.

Federal answer: No. The CFPB’s April 22, 2026 final rule eliminated effects-test liability from ECOA enforcement. Regulation B was amended to remove the disparate impact language. A borrower can’t bring a federal ECOA claim based solely on a disproportionate denial rate for their protected class.

Illinois answer, starting January 1, 2027: Yes. SB 3777 (Public Act 104-0744), signed July 31, 2026, amended the Illinois Human Rights Act to create an independent state-law standard prohibiting credit practices that produce discriminatory effects — regardless of intent.

For any lender with Illinois customers, those two answers need to coexist in your compliance program. The CFPB pulling back doesn’t change your Illinois Human Rights Act exposure.

What Illinois SB 3777 Actually Does

Before SB 3777, disparate impact claims under Illinois fair lending law generally tracked federal ECOA standards — meaning Illinois plaintiffs relied on the same legal theories and standards available at the federal level. When the CFPB eliminated disparate impact from Regulation B in April 2026, states that had imported the federal framework into their own standards lost some of that coverage.

Illinois decided not to let that happen.

SB 3777 amends the Illinois Human Rights Act in two ways that matter:

First, it explicitly prohibits the use of “facially neutral underwriting criteria or methodologies that produce discriminatory effects” in lending and credit card issuance. This is the disparate impact standard, stated in Illinois statute without relying on federal law to define it.

Second, it defines the terms and burden-shifting standard directly in the statute. The law doesn’t reference federal ECOA or Regulation B. It creates its own independent legal framework for what counts as disparate impact and how that claim works procedurally.

The practical consequence: Illinois’s fair lending obligations no longer rise and fall with how the CFPB interprets federal ECOA. Illinois’s General Assembly decided that the CFPB’s April 2026 rollback wouldn’t apply in Illinois — and built the legal architecture to ensure that’s true.

The Federal Backstory

To understand why SB 3777 matters, you need to understand what it’s responding to.

On April 22, 2026, the CFPB finalized a rule that amended Regulation B, the implementing regulation for ECOA, to remove the effects-test language that had provided the legal foundation for disparate impact claims under federal law. The rule was finalized after an extended comment period and represented a significant rollback from the Biden-era CFPB’s position on algorithmic lending.

Under the revised Regulation B, a lender can now make credit decisions using an algorithm that disproportionately denies credit to borrowers of a particular race, national origin, or other protected characteristic — and face no federal ECOA liability for that disparity, provided the decision wasn’t intentionally discriminatory. The discriminatory effects, standing alone, aren’t enough for a federal claim.

That change was significant for the fintech and AI-driven lending industry, which had faced ongoing uncertainty about disparate impact liability for machine learning models. ML models trained on historical credit data often exhibit disparate impact on protected classes even when neither race nor any proxy for race is an input variable — the patterns learned from data that was shaped by historical discrimination reproduce those patterns in predictions.

The CFPB’s April 2026 rule provided some clarity at the federal level. Illinois’s July 2026 law removed that clarity for any lender with Illinois borrowers.

Who’s Covered — and the Ambiguity About Fintechs

The Illinois Human Rights Act applies to “financial institutions” and “credit card issuers.” The Act doesn’t define “financial institution.”

That ambiguity is a live compliance question. Legal analysis from multiple firms suggests the term could extend beyond traditional depository institutions to:

• Nonbank mortgage lenders
• Marketplace lenders and personal loan platforms
• Buy now pay later providers
• Earned wage access platforms
• Business lending platforms with consumer-adjacent products

The argument for broad coverage is that the Illinois Human Rights Act’s lending provisions have historically been interpreted expansively. The argument for narrow coverage is that “financial institution” has a technical meaning under federal law that Illinois may have intended to import.

Until there’s regulatory guidance from the Illinois Department of Human Rights or case law clarifying the definition, any entity that extends credit to Illinois residents should conduct a coverage analysis. The safer assumption is covered unless you can affirmatively demonstrate you’re not.

Why AI Credit Models Are the Pressure Point

Manual underwriting with identifiable criteria can usually be analyzed for disparate impact and defended or modified. The exposure is manageable because the criteria are knowable.

AI and machine learning credit models are different. A complex model trained on historical credit data may produce disparate outcomes through pattern-matching that is difficult to fully explain — and that’s exactly the “facially neutral methodology that produces discriminatory effects” the Illinois law targets.

Before SB 3777, lenders using ML models could take some comfort from the uncertainty about federal ECOA disparate impact liability for algorithmic decisions. After January 1, 2027, Illinois residents who receive adverse credit decisions from an ML model that disproportionately affects their protected class have a potential state-law claim.

The Illinois burden-shifting framework applies:

1. Plaintiff demonstrates: The policy or methodology produces a discriminatory effect on a protected class.
2. Burden shifts to lender: The lender must demonstrate the practice is justified by a legitimate, nondiscriminatory interest — the business justification test.
3. If business justification shown: The plaintiff may still prevail by showing a less discriminatory alternative would achieve the same interest equally well.

For an ML credit model, satisfying the business justification test requires being able to articulate why the model’s methodology (including the features used and the way they interact) is the appropriate approach for the credit risk being assessed. “The model is optimized for predictive accuracy” is not a complete answer. “The model is optimized for predictive accuracy using these features, which are predictive of default risk and not proxies for protected characteristics, as validated by these tests” is closer to what the analysis needs to show.

The State Patchwork Problem

Illinois is not alone. California, New York, and New Jersey all maintain state-law disparate impact theories for lending that operate independently of federal ECOA. Illinois’s enactment adds the largest Midwestern state to that group.

The practical consequence for multi-state lenders: there is no single “federal compliance” posture for disparate impact that satisfies state obligations. A lender that operates only in states without independent disparate impact standards may find that federal ECOA’s April 2026 rollback genuinely reduces their exposure. A lender in Illinois, California, New York, New Jersey — or any combination — faces ongoing state-law disparate impact obligations regardless of what the CFPB does.

This creates a compliance program design challenge: the analysis, testing, and documentation needed to defend against a state disparate impact claim needs to be preserved and updated even as some federal obligations contract.

What Lenders with Illinois Borrowers Must Do Before January 2027

The effective date is January 1, 2027. That gives lenders roughly four months to address any gaps in their fair lending programs specifically for Illinois.

Disparate impact analysis on current models. If you’re using AI, ML, or any quantitative credit model for Illinois borrowers, run a disparate impact analysis specifically on your Illinois book. Look at approval and denial rates by race, national origin, sex, and other Illinois Human Rights Act protected classes. Document the results. If there are disparate outcomes, understand why — and whether you can demonstrate business justification.

Business justification documentation. For any model feature or underwriting criterion that contributes to disparate outcomes, prepare documentation of why that feature is predictive of credit risk and why it’s not a proxy for a protected characteristic. This documentation should exist before January 2027, not be constructed after a claim is filed.

Less discriminatory alternative analysis. The Illinois burden-shifting framework requires the lender to address whether a less discriminatory alternative would achieve the same credit risk prediction. This doesn’t require exhaustive testing of every possible model variant. It does require being able to show that you considered alternative approaches and that the alternatives you tested were less predictive, more costly, or otherwise inadequate for your credit risk needs.

Adverse action analysis. Adverse action notices for Illinois borrowers need to accurately state the reasons for denial. In an ML model context, that requires being able to extract the top factors driving a denial in terms a borrower can understand — regardless of what the model’s internal representation looks like. Illinois borrowers receiving an adverse action notice may cite disparate impact if they believe the stated reasons are pretextual.

Program update. Your fair lending compliance program should explicitly address Illinois SB 3777 — what it covers, how your program addresses it, and what testing cadence you’ve established. This is what a state examiner or plaintiff’s counsel will ask to see.

So What?

The CFPB’s April 2026 rollback on disparate impact felt like a resolution to a long-running legal uncertainty. For lenders in California, New York, New Jersey, and now Illinois, it wasn’t.

The state-federal divergence on disparate impact means that the practical compliance question — do we need to test our AI model for discriminatory effects? — doesn’t have a single national answer. It has a state-by-state answer that requires understanding which of your borrowers are in which states.

The AI Risk Assessment Template & Guide includes pre-built frameworks for bias testing and fair lending documentation specifically designed for ML credit models — covering disparate impact analysis, business justification documentation, and the less discriminatory alternative analysis that state-law claims will require. It’s built for the actual examiner question, not the question that existed before April 2026.

Illinois has four months before January 2027. That’s tight, but it’s workable if you start the analysis now.

Related reading: CFPB Finalizes ECOA Rule Eliminating Disparate Impact: What Your Fair Lending Program Must Do Now | The CFPB Gutted Federal Disparate Impact. Now AI-Driven Lenders Have a State Fair Lending Problem. | AI Bias Testing for Fair Lending: Methodologies Every Risk Team Needs