California's New Fair Lending Exam Law: What State-Chartered Banks, Credit Unions and Mortgage Licensees Should Have Ready

California's Fair Lending Examination Act (AB 801, Chapter 871 of 2026) requires the Department of Financial Protection and Innovation to examine state-chartered banks and credit unions that meet HMDA's loan-volume test, and every residential mortgage lender and servicer licensee, for compliance with federal and California anti-discrimination law in mortgage lending. Banks and credit unions get an exam at least once every four years. Licensees get it inside their existing exam, which must happen at least every 48 months.

Governor Newsom approved the bill on September 30, 2026. It is a non-urgency statute with no operative date of its own, so under Article IV, Section 8(c)(1) of the California Constitution it takes effect on January 1, 2027. Everything below is read from the chaptered bill text. Where the bill is silent, we say so rather than guess.

Who is covered

The Act adds two new chapters to the Financial Code and amends a third section, one for each type of institution DFPI supervises in mortgage lending.

State banks (new Financial Code section 1915) and state credit unions (new section 16910) are covered if they are "subject to the commissioner's examination authority" and meet at least one of the criteria in 12 CFR 1003.2(g)(1)(v). That paragraph of Regulation C is HMDA's loan-volume test: at least 25 closed-end mortgage loans in each of the two preceding calendar years, or at least 200 open-end lines of credit in each of those years, after the Regulation C exclusions.

The detail worth noticing is what the bill leaves out. Regulation C's full test for a depository institution also has an asset threshold, an office in an MSA, a first-lien home purchase loan and a federal nexus. AB 801 borrows only the volume paragraph. Read plainly, a state bank or credit union that clears 25 closed-end loans a year can be in scope for a California fair lending exam even if it falls under HMDA's asset threshold or has no office in an MSA, and so does not file a LAR. If that describes you, the data section below matters more than anything else in this article.

Mortgage lender and servicer licensees under the California Residential Mortgage Lending Act are covered through an amendment to section 50302. There is no volume test. DFPI already has to examine every licensee "at least once every 48 months," and the Act adds that "as part of the examination," the commissioner "shall examine the licensee for compliance with any nondiscrimination law applicable to mortgage lending."

Affiliates can be pulled in too. If a report or exam turns up "documented evidence of unlawful activity" between the institution and an affiliate, the affiliate can be examined on the same terms.

What a state fair lending exam will look at

The bill names five laws and uses the word "including," so the list is a floor:

• the federal Equal Credit Opportunity Act (15 U.S.C. 1691 et seq.)
• the federal Fair Housing Act (42 U.S.C. 3601 et seq.)
• California's Fair Employment and Housing Act (Government Code 12900 et seq.)
• the Holden Act (Health and Safety Code 35800 et seq.)
• the Unruh Civil Rights Act (Civil Code 51)

Two of those deserve a closer look. The first is FEHA. For banks and credit unions, the commissioner must "take appropriate steps to ensure correction of any violation of applicable nondiscrimination law, including a discriminatory effect under the California Fair Employment and Housing Act... that is not attributable to a legitimate business interest." That is disparate impact language, written into the exam statute. A state examiner is told to look at outcomes as well as intent and written policy.

The second is the Holden Act, California's anti-redlining law. Section 35810 bars a financial institution from discriminating in housing finance "due, in whole or in part, to the consideration of conditions, characteristics, or trends in the neighborhood or geographic area surrounding the housing accommodation," unless it can show the consideration was needed to avoid an unsafe and unsound practice. So geography is in scope by statute, alongside the applicant-level questions ECOA and the Fair Housing Act raise.

On procedure, the commissioner may examine books, records and documents and may question officers, directors, employees or agents under oath. Exams can also happen off schedule "for due cause, including receiving complaints or other evidence related to potential discrimination." The written findings go to the institution's principals, officers or directors and may be shared only with the institution, law enforcement and other regulators. The exam report is not a public record. Banks and credit unions pay a fee capped at the reasonable cost of the exam. A violation of any of the named laws becomes a violation of the Banking Law, the credit union law or the CRMLA, and DFPI can enforce it under those laws or under Division 24 of the Financial Code.

What the bill does not say: when the first four-year cycle starts, what an exam request letter will contain, or which statistical methods DFPI will use. When we checked DFPI's website on October 20, 2026, it had not published exam procedures or guidance under the Act. Until it does, the closest model for what examiners ask for is the federal one.

How it fits with federal fair lending exams

Federal regulators already examine for ECOA (Regulation B) and the Fair Housing Act. AB 801 lets DFPI lean on that work. The commissioner "may cooperate with any agency of the state or federal government," Fannie Mae or Freddie Mac, and "may accept an examination conducted by any of those entities in place of an examination by the commissioner," unless that exam "does not provide information necessary" for DFPI to do its job.

Read that carefully before assuming your last federal exam settles it. A federal exam covers ECOA and the Fair Housing Act. It does not test FEHA, the Holden Act or Unruh, so DFPI may decide it lacks the information it needs for those. The bill also tells DFPI to align the new exams with exams it already schedules and to "seek to enhance the efficiency and effectiveness of required examinations to reduce the total exam costs." An institution that "demonstrates full compliance" in its most recent exam under the new chapter may be exempted or examined less often. That relief only applies after a first exam under the Act, so nobody starts with it.

For a sense of how a federal fair lending exam runs, from the document request to the follow-up questions on a flagged disparity, see what a fair lending exam actually asks for. A state exam built on the same laws is likely to ask for much the same evidence, though the bill does not require it to.

The data you need ready

An examiner testing for discriminatory effect needs loan-level data that links outcomes to applicants and to locations. In practice that means:

• The HMDA LAR, if you file one. It already carries action taken, rate spread, denial reasons, credit score, debt-to-income and combined loan-to-value, applicant demographics, and the census tract for every reportable loan.
• What the LAR leaves out: underwriting exceptions, pricing overrides and any discretion a loan officer has over rate or fees. These are usually what explains a disparity by a legitimate factor, or fails to.
• Loans outside HMDA. Some mortgage products and applications fall outside Regulation C, and an institution that meets only the AB 801 volume test may not file a LAR at all. If you will be examined but do not report, you need a comparable loan-level file built from your loan system before the exam, not during it.
• A geocoded census tract, for every application, so the lending pattern can be read against tract income and minority population. That is the Holden Act question.

Where applicant race and ethnicity were not collected, as on most non-mortgage and some mortgage files, examiners often estimate them with a proxy. The usual method is Bayesian Improved Surname Geocoding. It estimates a probability, not a label, and it has known weak spots that are worth understanding before a proxy-based disparity shows up in a findings letter. We cover those in where BISG proxy methodology breaks down.

A pre-exam checklist

Our fair lending self-assessment guide walks through ten steps. Mapped to what AB 801 actually says, these are the ones to finish before a DFPI request arrives:

1. Confirm you are covered. Count closed-end and open-end originations for the last two calendar years against the 1003.2(g)(1)(v) thresholds. Licensees can skip this step.
2. Review policies against the California laws, not only the federal ones (step 1 of the guide). FEHA, the Holden Act and Unruh are now exam standards.
3. Assemble and scrub the loan data (step 2). Include the non-HMDA file if you need one, and geocode every record.
4. Run the statistical tests (step 3) on underwriting and pricing outcomes, controlling for the credit factors your policy uses.
5. Pull comparative files (step 4) for whatever the statistics flag, and review exceptions and overrides (step 5).
6. Check marketing and branch footprint by geography (step 6). This is where a Holden Act question would start.
7. Document findings and corrections (steps 8 and 9). Because the bill ties correction to effects "not attributable to a legitimate business interest," the record should say what the business interest is for any disparity that remains.

Where regression and redlining analysis fit

Disparate impact under FEHA and neighborhood discrimination under the Holden Act are two different tests, and each needs its own method.

Applicant-level effects are a regression question. Does a gap in denial rates or pricing between groups survive once the credit factors you actually use are held constant? Fair lending regression analysis: what it controls for and how to explain a pricing disparity covers the model, how to read its output, and how to document an explanation an examiner will accept.

Geographic effects are a redlining question. Compare your applications and originations in majority-minority and low- or moderate-income tracts with what peer lenders achieved in the same markets, and look for gaps in branch and marketing coverage. Our redlining analysis page explains how that review is put together.

Where Comply fits

Comply Fair Lending runs the regression and BISG proxy analysis described above on your own loan data, and the redlining review draws on Comply Fair Lending, Peer-2-Peer, Mapping and DataMine together. If you would rather not run the software on your own servers, RATA Hosted Services hosts every Comply module for you. The point is to know what DFPI will see in your data before it asks.

Schedule your free demo today

Sources

• AB-801 Nondiscrimination (2025-2026), chaptered text
  California Legislative Information. Chapter 871, Statutes of 2026; approved by the Governor and filed with the Secretary of State September 30, 2026. Retrieved October 20, 2026.
• AB-801 bill status
  California Legislative Information. Non-urgency measure. Retrieved October 20, 2026.
• California Constitution, Article IV, Section 8(c)
  Effective date of statutes enacted at a regular session. Retrieved October 20, 2026.
• 12 CFR 1003.2(g), Regulation C definition of financial institution
  Electronic Code of Federal Regulations. Retrieved October 20, 2026.
• Health and Safety Code section 35810 (Holden Act)
  California Legislative Information. Retrieved October 20, 2026.
• Department of Financial Protection and Innovation
  Checked October 20, 2026 for guidance or exam procedures under the Act; none published.

Photo by Robert So on Pexels.