The CFPB Dropped Disparate Impact. State AGs Didn't.

Five months ago, the CFPB quietly dismantled one of the central pillars of fair lending enforcement. The April 22, 2026 final rule removed the “effects test” from Regulation B and determined that ECOA does not authorize disparate-impact liability.

If you read that as permission to relax your fair lending program, you read it wrong.

State attorneys general did not get the memo — or rather, they got it and decided it didn’t apply to them. Since April, state-level fair lending enforcement has intensified, coordinated, and expanded its analytical toolkit. The banks that scaled back their fair lending infrastructure because the federal standard loosened are now facing something more complicated: a patchwork of state standards, some stricter than the old federal baseline, enforced by agencies that are better resourced and better coordinated than they were 18 months ago.

TL;DR

• The CFPB’s April 2026 final rule removed disparate impact liability under ECOA — federal enforcement now focuses on intentional discrimination
• State AGs retain independent authority to pursue disparate impact claims under state laws and are actively exercising it
• Former CFPB Director Chopra is advising state AG coalitions, enabling coordinated enforcement using shared data
• AI-driven lending models generating statistically disparate outcomes remain exposed at the state level
• The DOJ continues federal redlining enforcement under the Fair Housing Act, unchanged by the ECOA rule
• Your fair lending program needs to account for your state footprint, not just federal standards

What the CFPB Actually Did

The legal history here matters because it shapes how state enforcement authority works.

The Equal Credit Opportunity Act prohibits discrimination against credit applicants based on race, color, religion, national origin, sex, age, marital status, familial status, or the fact that income derives from public assistance. For decades, the CFPB and its predecessor agencies interpreted ECOA as incorporating a disparate impact theory — meaning lenders could be held liable not only for intentional discrimination but for facially neutral policies that produced statistically disparate outcomes for protected classes.

That interpretation came under increasing legal pressure after Inclusive Communities (2015) and a series of circuit court decisions that questioned whether private disparate impact claims were cognizable under ECOA. The CFPB’s April 2026 final rule resolved the question administratively: the agency determined that ECOA does not, as a matter of statutory interpretation, authorize disparate-impact claims. Regulation B was revised to remove the effects test accordingly.

The immediate practical effect for federal examination purposes: CFPB examiners are no longer evaluating lenders for facially neutral policies that produce disparate outcomes. Intentional discrimination remains illegal and remains an enforcement priority. The evidentiary burden on the agency shifted; the target of enforcement narrowed.

What did not change: state authority to enforce state fair lending and civil rights statutes. And those statutes, in most states, still carry robust disparate impact liability.

The State Enforcement Picture

States enforce fair lending through several mechanisms: state civil rights statutes, state consumer protection laws, state banking laws, and their own banking examination authority. The CFPB’s interpretation of a federal statute has no binding effect on state enforcement authority under state law.

Several states moved quickly after April to signal that their own enforcement priorities had not changed. Illinois enacted SB3777 to explicitly preserve and strengthen disparate impact claims against lenders under state law — a direct legislative response to the CFPB rule. California’s DFPI issued guidance clarifying that disparate impact analysis remains part of state fair lending examinations. New York’s DFS has long maintained independent fair lending enforcement apparatus that the CFPB rule does not affect.

But the more significant development is coordination. Former CFPB Director Rohit Chopra, after leaving the agency, has been actively working with state attorneys general on coordinated enforcement strategies. Several AG coalitions have formalized data-sharing agreements — meaning a lending pattern examined by one state’s AG can be shared with counterparts in other states for coordinated enforcement.

That coordination changes the practical exposure calculus for multi-state lenders significantly. A disparate impact pattern that previously required each state to develop its own case independently can now result in a coordinated multi-state action. The enforcement leverage is materially higher.

What’s in the Crosshairs

State enforcers have been direct about their enforcement priorities in the post-CFPB landscape. Three areas come up consistently.

AI credit decisioning models. Lenders using machine learning models in underwriting, pricing, or marketing selection face heightened scrutiny at the state level. The concern is straightforward: models optimized for predictive accuracy can embed historical disparities from training data, producing statistically disparate outcomes for protected classes even without any discriminatory intent. State examiners and AG investigators are specifically requesting model documentation, disparity testing results, and adverse action explanation methodologies for AI systems.

This isn’t a new concern — it predates the CFPB rule change. But the rule change removed a federal counterweight that had structured how lenders engaged with these questions. With the federal expectation gone, state-level scrutiny has intensified to fill the space.

Pricing discretion and algorithmic pricing. Pricing algorithms that incorporate geographic or demographic proxies face disparate impact analysis at the state level. This includes mortgage pricing models, small business lending pricing, and increasingly, buy-now-pay-later and consumer installment loan pricing. State examiners have been asking for pricing disparity analyses as part of routine examinations in California, New York, Illinois, and Massachusetts.

Redlining through digital marketing and product design. Modern redlining doesn’t look like a map with a red line drawn through it. It looks like targeted marketing that concentrates financial product promotion in majority-white geographies, digital product designs that presuppose infrastructure (credit history, bank accounts, broadband) that correlates with race, and branch network decisions that effectively exclude majority-minority communities. State enforcers have adapted their analytical frameworks accordingly.

The FTC’s enforcement surge in financial services has also added a parallel enforcement axis — unfair or deceptive practices claims can be layered with fair lending analysis in ways that multiply exposure.

The Federal Picture Isn’t as Simple as You Think

Even at the federal level, the CFPB rule change is narrower than some compliance teams have concluded.

The Department of Justice’s Civil Rights Division fair lending enforcement continues unchanged. DOJ investigations proceed under the Fair Housing Act, which has its own disparate impact theory that the CFPB’s ECOA interpretation does not touch. Redlining cases, in particular, tend to be brought under FHA as well as ECOA — removing ECOA disparate impact liability doesn’t make a redlining investigation go away.

The CFPB Accountability Reform Act created additional political and procedural complexity around CFPB enforcement, but federal banking examiners at the OCC, Fed, and FDIC continue to conduct fair lending examinations. Those examinations still include disparate impact analysis in their scope — the OCC and Fed did not adopt the CFPB’s ECOA interpretation as the standard for their own supervisory work.

Additionally, consent orders with disparate impact commitments already entered into court are not affected by the rule change. Institutions operating under existing consent orders with disparate impact remediation requirements remain bound by those commitments.

What Your Fair Lending Program Needs Now

The practical compliance response to this environment is more nuanced than “we can relax on disparate impact.” It requires mapping your actual exposure by jurisdiction and by enforcement mechanism.

Map your state footprint. Where do you do business? Which states have state fair lending statutes with disparate impact provisions? Which state banking regulators conduct independent fair lending examinations? Which state AG offices have been active in financial services enforcement? This analysis should produce a jurisdiction-by-jurisdiction risk map that tells you where your disparate impact exposure is highest.

Maintain disparity testing regardless of federal expectations. Disparity analysis on your credit decisions, pricing, and marketing has value beyond regulatory compliance — it surfaces model performance issues and concentration risks that matter for business reasons. The testing infrastructure your fair lending program built for ECOA compliance is still necessary for state examination and litigation defense. Don’t dismantle it because the federal examination scope narrowed.

Document your AI model governance for the right audience. State regulators asking for AI model documentation want to understand whether your underwriting or pricing model produces disparate outcomes, what testing you did, what you found, and what you did about it. The documentation standards that satisfy federal examination may need to be more detailed for state examination — particularly in California, New York, and Illinois. Model cards, disparity test results, and adverse action explanation methodologies should be maintained for every AI system influencing credit decisions.

Build your explanation capability. Adverse action notices for AI-driven credit decisions face scrutiny from state regulators on whether the explanations are specific, accurate, and actionable. The CFPB loosening federal expectations doesn’t reduce state expectations on explanation quality. If your AI models can’t generate accurate, legally compliant adverse action explanations, that’s a state examination risk and a litigation risk regardless of federal standards.

Watch the consent order landscape. Multi-state enforcement actions are more likely in this environment than before. A single coordinated investigation involving five state AG offices and a referral to DOJ is a qualitatively different legal risk than five independent state examinations. Stay current on what other institutions are being investigated for — the pattern of enforcement tells you what’s coming next.

The Temptation to Underfund Fair Lending Compliance

When federal enforcement expectations narrow, the business case for maintaining the same level of fair lending compliance investment weakens. That’s a predictable institutional response. It’s also a trap.

The institutions that cut fair lending staff, reduced disparity testing frequency, or deprioritized model documentation after April are the institutions most exposed to the coordinated state enforcement environment that’s developed since then. The cost of remediating a state AG investigation — discovery, legal exposure, remediation programs, potential monetary relief — is substantially higher than the cost of maintaining the compliance infrastructure.

State enforcement of fair lending is now structurally more vigorous than it was 18 months ago. The federal standard narrowed. The state standard didn’t. Programs built for the old combined federal-plus-state exposure are now, in many cases, actually more necessary than they were before the CFPB rule changed — because the state enforcement piece is carrying more of the weight.

So What?

The CFPB’s ECOA rule change was real and it mattered. Federal disparate impact exposure under ECOA is reduced. But the assumption that overall fair lending risk declined is incorrect in most multi-state lending contexts.

State AGs are coordinated, analytically sophisticated, and explicitly filling the space the CFPB vacated. AI-driven lending decisions that produce disparate outcomes remain exposed at the state level. The DOJ’s redlining enforcement continues under the Fair Housing Act. Existing consent orders are unchanged.

The right response is to understand exactly where your exposure lives — by jurisdiction, by enforcement mechanism, and by product line — and calibrate your program accordingly. Not to reduce investment across the board, but to redeploy it where the actual risk has migrated.

Fair lending compliance didn’t get easier in April 2026. It got more complicated.

Tracking fair lending KRIs across a multi-state lending program? The KRI Library — Key Risk Indicators includes pre-built fair lending and compliance risk indicators designed for financial services institutions.

External sources:

• CFPB’s Final Rule Recalibrates Fair Lending Enforcement — Consumer Finance Monitor
• CFPB Final Rule Reshapes Fair Lending Enforcement Under ECOA — National Mortgage Professional
• Fair Lending in 2026: Why Quieter Doesn’t Mean Calm — Ncontracts
• DOJ Fair Lending Enforcement — Department of Justice Civil Rights Division
• September 2026 Regulatory Update: Fair Lending Rulings — Ncontracts