All posts by Synergy

Mortgage Industry AI Copilots: What Has to Be in the File Now

One vendor is now running more than 3,500 new loans a week through an AI layer — an annualized pace close to 200,000 files. Another has logged more than four million verified AI tasks on production loans, at a claimed 99 percent accuracy on critical fields including signatures, notary stamps and dollar amounts. These are not pilots. They are production systems, and somewhere in the country a Fannie Mae seller is one phone call away from being asked to describe every one of them.

That call is not hypothetical. Lender Letter LL-2026-04 took effect on August 6, 2026, and it makes AI governance a written eligibility requirement rather than a best practice. The mortgage industry AI copilots question is no longer whether you should deploy them. It is whether your organization can produce, on request, the record that proves what you deployed and what you did to control it.

What is actually running in production right now

Four examples, because they show the range of what the label covers.

• Blend Autopilot — an agentic intelligence layer for home lending. The tenth lender signed on in October 2026, including Whatcom Educational Credit Union and Gate City Bank. It works on more than 3,500 new loans per week, roughly 200,000 annualized, six months after it started on live loans. Reported results: 10 to 15 percent higher pull-through, cycle times shortened by two to four days, and an average of 4.5 hours of fulfillment work automated per loan. The work spans document review, income review, borrower follow-ups, pre-underwriting, and closing quality control. Onity Mortgage, which serves 1.4 million customers, is an early adopter that started with the pre-underwriting agent.

• Areal Copilot Agent — more than 4 million verified tasks on production loans as of June 16, 2026. On a single loan, individual agents complete anywhere from 10 to more than 100 discrete tasks: document indexing, signature and notary-stamp checks, fee comparisons, loan-amount validation, exception drafting, LOS updates. The company reports 99 percent accuracy on critical fields across more than 1,500 document types and more than 4,000 data points per loan, with 2X or greater origination and closing throughput on the same headcount. Lenders author agents in near-natural-language prompts, scope each agent’s access to specific documents, LOS fields, contacts and tools, and retain a full traceable audit trail on every action.

• eLend — building its own loan origination system with a staff of AI engineers rather than database administrators. Agents are in production reviewing powers of attorney, trust documents, homeowner’s insurance declaration pages, flood certifications and title commitments. The company’s stated position is unambiguous: “We’re never, ever, ever going to have AI just signing off on loans or making decisions because legally you can’t … That’s why it goes to our underwriter. They have a checklist they run against.”

• Skipton Building Society in the UK — by July 2026, 80 percent of a workforce of roughly 2,500 was using Microsoft Copilot Enterprise. A change cannot be rolled out beyond 10 to 15 people without sign-off from the relevant team. The stated principle: “You can all have approved tools, but there are rules around how you use them. And we will monitor.”

Read those four together and the pattern is clear. The mortgage industry AI copilots in production are not experiments waiting to be governed. They are running on live borrower files right now, and the governance question has moved from the model to the record. The industry has adopted the copilot model rather than the replacement model. AI does the work; a named human signs. That is a sound legal posture. It is also an operational one, because it means the compliance burden has moved from “did the human check it” to “can we show what the machine did and what the human did about it.”

The requirement that is already in force

Fannie Mae issued LL-2026-04 on April 8, 2026, made it effective 120 days later on August 6, and updated it on October 21, 2026. Any Fannie Mae seller or servicer that uses AI or machine learning in origination or servicing must operate under a documented governance program. It does not carve out internally built tools versus vendor systems.

The letter requires six things.

1. Policies and procedures covering the development, implementation, use and maintenance of any AI or ML system the lender relies on, plus a method for measuring and managing the associated risk.

2. Specific qualities in those policies — transparent, communicated to the appropriate personnel, incorporating the characteristics of trustworthy and ethical AI, reflecting an understanding of applicable legal and regulatory requirements, and aligned with the firm’s stated risk tolerance.

3. At least annual review of the policies for compliance with law and consistency with industry practice, with the review itself documented.

4. Vendor and subcontractor governance that is no less protective than the lender’s own standards. In practice, the third party running your fraud screen, document classifier or borrower-facing chat-bot has to meet your controls, not just its own.

5. Disclosure on request. On request, promptly tell Fannie Mae the types of AI and ML in use, the purpose and manner of that use, the safeguards in place to mitigate AI and ML risk, and any other information required. There is no annual filing, but there is a standing obligation to be ready to answer.

6. Consistency with the Information Security and Business Resiliency Supplement, which the letter cross-references directly — so an AI use case that touches the security Supplement is assessed against both at once.

Freddie Mac updated its Seller/Servicer Guide in March 2026 to require governance frameworks for AI and machine learning. Between the two, the AI governance question is now a contractual eligibility matter for a large share of the market. It is not a best-practice document you can adopt later, and it is not a thing that only applies if you built the tool yourself.

The framework that landed with it

About a year ago, the MBA’s Residential Board of Governors made responsible AI guardrails a 2026 priority. MISMO already had the infrastructure to act on it — an AI community of practice with more than 100 members across the industry. In under six months, they published FRAME, the Framework for Responsible AI in Mortgage Ecosystems.

FRAME gives lenders two concrete things. The first is a policy and procedures template, for firms that have no written AI policy at all. The second is a use-case inventory tool, which is generally described as the more important of the two, for a reason that matters operationally: one vendor can carry three or four distinct AI use cases, and each one gets its own line item, its own owner, and its own risk tier.

That is also where the industry guidance converges on a rule worth adopting now. Anything borrower-facing should be enterprise-only — vetted, under contract, and with the lender on the hook for it. And the risk tiers should carry service level agreements: low-risk use cases get an expedited approval window, high-risk ones get a consistent testing program that samples 5 to 10 percent of chat-bot conversations or email outreach, the same way quality control already samples loans. Consistent tests produce consistent turnaround. That is the whole point of having a sampling plan rather than a rule.

The economics nobody wants to say out loud

The adoption data is encouraging. The profitability data is not, and the two are connected.

A BCG, MBA and AARMR survey fielded between April and July 2026, published in September, found that mortgage lenders trail other financial institutions — insurance and banking — in AI development, and that the industry is still early in AI maturity. The survey sample was small at 31 lenders, so read it as a direction rather than a measurement.

Meanwhile, HousingWire reported that in the first quarter of 2026 the average cost to originate a loan rose to $11,898, while lenders earned $727, or 16 basis points, in pre-tax production profit. The analysis is the important part. Most AI deployments improve individual tasks inside the existing workflow. Layering AI onto the same fragmented software stack accelerates work at specific points while leaving the underlying operating model intact, and accelerating work without changing the operating model does not change the economics. Buying AI is easy. Removing hand-offs is hard, because it means reconsidering workflows, roles, data ownership, controls, incentives and how productivity is measured.

The first wave of mortgage AI has produced copilots and incremental productivity. That is a legitimate outcome, and it is what most of the market has bought. But it means the return case for AI is not a margin story that closes by itself. It is a story that requires the governance work, because the governance work is what removes the hand-offs.

Where the exposure actually is

This is the part that changes how a lender should think about its own AI program. Any loan that defaults gets quality controlled by the agencies. Aggregators do the same. If a default can be tied to a defect that was caused or amplified by AI, the exposure is an indemnification or a repurchase. Reps and warranties roll downhill, and they roll downhill through the vendors you bought from.

The mitigation is not to remove the AI. It is to use it to check your own work first and break that loop, and to be able to prove you did. Which means the audit trail is not a nice-to-have feature of a platform. It is the artifact that stands between a default and a repurchase.

Two practical consequences follow. First, if a vendor cannot show you a traceable log of what its agent did to your file, you have bought an ungoverned actor inside your loan process. Second, borrower-facing AI is the highest-risk category in the stack, because the output of a chat-bot or email drafting tool can end up in a file, in an adverse action analysis, or in a complaint — which is exactly the evidence set that the amended Regulation B still examines.

The AI inventory: the artifact lenders are missing

If you do one thing from this article, build the inventory. It is the single most useful artifact when a GSE asks what you use and why, and it is consistently the item lenders do not have.

It is also, unusually, not that much work. The industry guidance is straightforward: ask loan officers and processors to list every AI use case they rely on, from up-funnel marketing through post-close. Not the tools you bought. The things the tools do. “Document review,” “income verification,” “borrower follow-up,” “exception drafting,” “closing QC.” Then add each one to your inventory, and give it an owner, a business purpose, a risk tier, and a testing method.

When that inventory exists, three other things become easy. The vendor questionnaire becomes a conversation instead of a questionnaire. The annual policy review has something to review. And the disclosure package for Fannie Mae can be assembled from artifacts you already maintain instead of from a scramble.

Seven things to have in place before January 1

1. A written AI and ML governance policy that defines what counts as AI at your firm, assigns ownership, states your risk tolerance, and describes how you develop, deploy, monitor and retire models. Review it annually and record the date and outcome of that review, because the cadence has to be provable.

2. A complete use-case inventory covering internal tools and vendor tools, each with a documented business purpose, a named owner, and a link to an origination or servicing activity.

3. Vendor and subcontractor governance that reaches your contracts and due-diligence questionnaires, with the right to obtain from any AI vendor the same disclosure information a GSE might request from you: types, purpose, manner of use, safeguards.

4. Monitoring and testing logs. Track model performance over time, watch for drift and for bias, and log what you found. An audit trail with no recorded outcomes is an audit trail with no evidence.

5. A prepared disclosure package — AI and ML types in use, purposes, manner of use, safeguards, governance documentation — assembled and current, because both GSEs can ask and requests do not come with a deadline.

6. Senior accountability with a named executive who owns the program. A governance program without a name attached to it is a document, not a control.

7. A sampling plan for borrower-facing AI that samples 5 to 10 percent of conversations and outbound messages the way quality control samples loans, with published service level agreements for both expedited and high-risk use cases.

None of these seven items is technically difficult. All of them are boring. That combination is the defining feature of this moment in mortgage AI: the tooling is ahead of the governance, and the institutions that get the governance right early will be the ones still operating when the tooling is ordinary.

If you want your AI inventory pressure-tested against what a GSE might actually ask, Synergy’s compliance review service works from your vendor contracts and your sampling plan rather than from a template. And if it is faster to start with one artifact, request a review through Synergy and we will tell you which of the seven you can prove today and which one is going to be asked about first.

The Mortgage Loan Processor in 2027: What Volume Collapse Doesn’t Fix

There were fewer loans in October. There were also fewer experienced people processing them (mortgage loan processor), fewer lenders competing for the same ones, and fewer dollars on the balance sheets of the companies that fund them. Every one of those is a market condition. None of them is a compliance control, and none of them reduces what a single file has to prove.

The volume numbers are stark enough to lead with. Total rate-lock volume fell 19 percent month over month in October and 38 percent year over year, according to Optimal Blue’s Market Advantage report released November 10. Purchase volume fell 18 percent month over month and 17 percent year over year, and still accounted for 84 percent of October locks, leaving refinance at 16 percent. Adjustable-rate mortgages reached 17 percent of lock volume, the highest share in a data series that goes back to 2019.

So the environment is thinner in every direction. The question for a mortgage loan processor, a processing company, and the compliance program that supports them is what a thinner environment does to the file. It does not make the file simpler. It makes the denominator smaller, the staffing less experienced, and the consequences of a single defect heavier, because each loan is now a larger share of the book.

What the market did in October

Rates eased slightly but did not come back. The Freddie Mac Primary Mortgage Market Survey put the average 30-year fixed rate at 7.24 percent for the week ending October 29, down from 7.32 percent the week before and 7.40 percent earlier in October. A year earlier the 30-year averaged 6.17 percent. The 15-year fixed fell to 6.60 percent, against 5.41 percent a year earlier.

And demand did not improve in a way that helped anyone. According to the National Association of Home Builders, 46 percent of builders reduced home prices in November, the highest share recorded since the pandemic, while 68 percent offered incentives such as mortgage rate buy-downs and closing-cost assistance. When a builder is buying down a rate to make the house affordable, that buy-down is a document in your file with a funding date, a compliance review, and a disclosure consequence.

The refinance market, the part of the business that carried volume through the last two rate cycles, is down to 16 percent of locks. That is the part of production that was most automated and most sensitive to a file being complete. It is the part that got smaller first.

The cost of getting through it

The industry is not absorbing this with a stable cost base. It is absorbing it with consolidation and layoffs.

• Mortgage Solutions Financial, a Colorado lender and servicer holding about $1.9 billion in originations, is moving its entire origination division and sales team to Fairway Independent Mortgage Corporation effective November 30, 2026, and is shifting focus from origination to servicing.

• Pennymac is expanding layoffs to cover $160 million in annualized operating expenses, after originations and acquisitions fell 30 percent.

• Rocket Companies has said its mortgage business will end 2026 and 2027 with originations below industry forecasts, and expects 2027 below $2 trillion with fourth-quarter refinances nearly halving.

• Stratmor expects to see 46 lender M&A deals by the end of the year.

• Fitch downgraded four independent mortgage banks over high non-funding debt levels, warning that a smaller 2027 market will limit earnings upside.

• UWM is holding gain-on-sale at 115 to 130 basis points while some wholesale rivals dropped to 70 to 80 basis points to win loans.

Every one of those events moves files. A branch that closes sends its pipeline. A servicing transition re-papers the file. A bankruptcy changes who is holding the documentation. A lender that cannot fund a loan at 80 basis points takes a smaller warehouse line and services it harder. The processing work does not shrink in proportion to the volume; it changes character, and the changes land on people who are new to the workflow.

The quality control number is the one to watch

ACES Quality Management released its quarterly Mortgage QC Industry Trends report on November 10. The headline reads well: the overall critical defect rate fell to 1.64 percent in the second quarter of 2026, a 4.1 percent drop from the first quarter, and the first decline between the first and second quarters since 2023.

Then read the rest of the sentence: the rate remained 8.6 percent higher than a year earlier.

That combination is the story of 2026, and it deserves more attention than it is getting. The sequential improvement is real, and it is the kind of number that gets quoted in a board deck without the second clause. The year-over-year level is the one that describes your file. A lender that was running a 1.5 percent critical defect rate in 2025 is running something meaningfully worse than that today, in a market where volume is down more than a third, where staffing has been cut, and where a single defect has a larger effect on the book.

This is also the number to watch as 2027 reporting lands. If the sequential decline continues and the year-over-year gap closes, that is a genuine improvement and it will be visible. If the gap stays wide, then a lower volume is not producing a cleaner file, and the assumption baked into most 2027 staffing plans is wrong.

What a mortgage loan processor is carrying into January

Beyond the market, there is a short list of specific 2026 policy that lands on a processing desk right now.

The appraisal data set changed on November 2

The Uniform Appraisal Dataset 3.6 policy supplement and the redesigned Uniform Residential Appraisal Report took effect November 2, 2026. Fannie Mae Lender Letter LL-2026-08, issued September 30, grants a temporary policy exception for lenders unable to implement in time. The November date did not move. If your files were delivered on the old format, you are now holding two populations of appraisals and two sets of form checks.

Adjustable-rate mortgages are a quarter of the lock volume

At 17 percent of lock volume, ARMs are not a niche product any more. That is a materially different underwriting and disclosure path — note rate, caps, payment shock language, the ATR analysis, and a different set of conditions from a fixed loan. A processing operation that treated ARMs as an exception is now running an exception at scale.

The 2027 loan limit is moving and the delivery date governs

This is the live one. As of November 17, 2026, FHFA had not announced the 2027 conforming loan limits. The 2026 figures remain in force: $832,750 for a one-unit property in most of the country and $1,249,125 in high-cost areas. FHFA has announced each year’s limits between November 25 and November 30 for the last five years, so the announcement is days away. HUD’s FHA limits follow in December.

Meanwhile, more than 50 lenders are already originating to unofficial 2027 limits, in three competing tiers: $845,000, $847,440 and $850,000. Those are lender business policies, not federal limits, and a loan written above the official limit cannot be sold to Fannie Mae or Freddie Mac until an eligible execution exists. The lender holds the exposure in the meantime.

The operational detail that causes the most rework is which year applies. Fannie Mae applies the loan limit based on the original loan amount and the area where the property is located at the time the loan is delivered, per Selling Guide B2-1.5-01. A loan that closes in late December 2026 and is delivered in January 2027 may be tested against the 2027 number. If the file was underwritten to the 2026 limit and the 2027 limit lands lower than what you qualified the borrower on, that is a re-underwrite, not a clerical fix.

Delinquencies are climbing again

Fannie Mae and Freddie Mac-backed delinquencies reversed course from early summer and are climbing in tandem with the ongoing rise in inflation, according to National Mortgage News. Servicing exposure, loss mitigation, and the ability to produce a complete quality control package on a delinquent file are all operating on a deteriorating base.

Six things to fix in the next sixty days

1. Write down the loan limit rule you are actually applying, and when. One page: which limit applies on origination date, which applies on delivery date, what your lender does with a file between the two, and what happens to a borrower who qualified at $845,000 if FHFA lands lower. Every processing operation has this rule somewhere, and almost nobody has it written.

2. Separate the appraisal populations. Tag every open and un-delivered file by appraisal form version. You are going to need to answer “which format was this?” for every file you deliver for the next two years, and the answer should not require opening the file.

3. Re-classify ARM processing as a core workflow, not an exception path. If 17 percent of locks are ARMs, the condition set, the disclosure set and the training for the loan processor handling them should look like the fixed-loan path, not like a special approval.

4. Reconcile your defect rate against last year, not against last quarter. Take the ACES figure and the equivalent figure from Q2 2025 and put them side by side for the board. Then break your own internal critical defect rate down by defect source, because the aggregate tells you whether to look at people or at systems.

5. Document what happens to files that move during a transition. Branch sale, servicing transfer, lender acquisition, employee move to a competing shop. For each, you need to know where the originals live, who is allowed to pull them, and how the release was authorized. This is the documentation that is missing when a lender needs it during a repurchase discussion.

6. Reconcile your AI use against your QC sample plan. If a copilot or an agent touches the file — document indexing, income review, exception routing, closing quality checks — then the sampling plan has to say so. Sampling the wrong population is the most expensive mistake in this list, because it looks rigorous while testing nothing.

The one-paragraph version

Volume is down 38 percent. That is a business problem. The critical defect rate is still up 8.6 percent year over year, four lenders were downgraded in a single week, a fifth of locks are adjustable-rate, your appraisal format changed on November 2, and your loan limit is about to change on a date that depends on when the file delivers rather than when the borrower signs. Those are control problems. A mortgage loan processor who treats 2027 as a smaller version of 2026 is going to be fine on volume and wrong on the file.

If you want the operational side pressure-tested — the file flow, the condition sets, the QC sample, the transition documentation — Synergy’s mortgage loan fulfillment support works inside your existing operation rather than around it. And if you would rather start with one hard question about your own file, our closing and QC review service takes a sample and tells you what we would pull.

CFPB Compliance in 2027: What Actually Changed

Most of the commentary on CFPB compliance in 2026 has been about the Bureau closing its books. That is the least useful thing to plan around. The Bureau published its Semi-Annual Report to Congress on October 27, and the headline number is stark: 1,327 of 1,552 supervisory actions closed, about 86 percent, and scheduled examinations cut by roughly half. Sixty-six of eighty open examinations were closed. Sixteen pending final orders were terminated.

But a rule book does not shrink because an agency stopped reading it. On April 22, 2026, the CFPB issued a final rule amending Regulation B, the regulation implementing the Equal Credit Opportunity Act. It became effective on July 21, 2026. It is currently being litigated in the District of Massachusetts, and it is effective today.

This is what a 2027 CFPB compliance plan actually has to account for: a smaller examination footprint, a rewritten standard for what counts as discrimination, and a live legal challenge that means the standard could move again inside your policy horizon. Here is what each of those pieces changes.

What the Bureau closed, and what it kept

The Semi-Annual Report covers January through June 2026, with a reporting period running from July 1, 2025 through June 30, 2026. The enforcement and supervision numbers describe a Bureau resolving its own backlog rather than adding to it.

• One new public enforcement action initiated. Nine resolved. Sixteen pending final orders terminated, in all but one case after the redress and penalties had already been paid. One action dismissed or withdrawn from as plaintiff.

• Approximately 40 percent of pending investigations closed after review against the revised priorities.

• 1,327 of 1,552 supervisory actions open as of July 1, 2025 closed — about 86 percent. Closures reflected either institutions addressing the finding or the finding no longer matching Bureau priorities.

• All eighty open examinations reassessed. Sixty-six closed. Fourteen continue, focused on correcting violations, addressing compliance weaknesses, and providing relief to consumers.

• Scheduled examinations reduced by approximately half, with attention shifting toward depository institutions, matters the Bureau characterizes as actual consumer fraud, and areas inside its statutory authority.

• 8.27 million complaints received; 7.50 million sent to companies for review and response.

The Bureau also set out four enforcement principles, issued in May 2026: address actual consumer harm, respect due process, collaborate with institutions to resolve problems voluntarily, and avoid duplicative enforcement already conducted by state or other federal regulators. That last one is aimed squarely at the states, and it changes what a state examination looks like on your calendar.

Two priorities in the report should make an originator sit up. The first is mortgage-related matters — mortgage origination and servicing remain named supervisory priorities even as the overall footprint shrinks. A smaller program pointed at the same place is still pointed at you. The second is debanking, which the Bureau has elevated alongside service member redress and FCRA and Regulation V data-furnishing violations. If your institution has been reviewing account closures, that is now a named area of focus rather than an afterthought.

The rule that is already in force

The April 22 final rule, published in the Federal Register at 91 FR 21621 and effective July 21, 2026, made three changes. All three are finalized. None is a proposal waiting for comment.

One: disparate impact is gone from ECOA

The Bureau removed the effects test and stated affirmatively that ECOA does not authorize disparate-impact liability. A facially neutral policy that produces a disproportionate exclusion no longer supports a claim on its own. Liability survives only where neutral criteria function as proxies for protected characteristics and are applied with the intent to advantage or disadvantage a protected group. Intent, not disparity, is now the load-bearing element.

Two: the discouragement provision narrowed sharply

Regulation B section 1002.4(b) now prohibits only statements of intent to discriminate. The Bureau limited the prohibition to spoken or written words or visual images, and stated that encouragement directed toward one group does not constitute discouragement toward another. It also drew a line around evidence: branch locations, advertising locations, and community events sit outside the prohibition. A lender can no longer be found to have discouraged an applicant based on where it placed a branch or ran an ad.

Race, color, national origin, and sex are barred as common characteristics in a special purpose credit program. For-profit creditors running SPCPs face additional requirements. This is the piece most likely to be missed by lenders that do not run SPCPs themselves and most likely to cause friction for lenders that do.

The rule is in court. That is not the same as safe.

On October 29, 2026, the National Urban League and the Massachusetts Association of Community Development Corporations filed suit against the CFPB and Acting Director Mark Paoletta in the District of Massachusetts, in National Urban League v. Consumer Financial Protection Bureau, No. 1:26-cv-14891-NMG. The National Association for the Advancement of Colored People Legal Defense Fund and the National Consumer Law Center filed on their behalf. The plaintiffs ask for declaratory and injunctive relief, including vacating the rule.

The complaint alleges that the Bureau acted arbitrarily and capriciously under the Administrative Procedure Act, that it failed to provide reasoned explanations, respond meaningfully to comments, consider reliance interests, or assess costs and benefits. It alleges the comment period was too short at thirty-two days, that the Regulatory Flexibility Act and the Small Business Regulatory Enforcement Fairness Act were not satisfied, and that the Dodd-Frank Act consultation requirement was not met. It alleges the acting director lacked authority to promulgate the rule at all.

The number that should stop you is the comment count. The Bureau received approximately 64,500 comments on the proposed rule, about ninety percent of them opposed, and changed nothing. The plaintiffs say the rule is void.

Here is the part that matters operationally: the rule is effective and has been since July 21. A pending challenge is not a stay. Until a court says otherwise, every lender is operating under the amended standard, building files under it, training staff on it, and writing adverse action reasons against it. The realistic outcome is not that the rule vanishes next month. It is that the litigation produces delay, and possibly a remand that sends the question back for a different rule making.

Which means the correct planning assumption is not “wait and see.” It is that the narrower standard holds through 2027 and that you still need the records to defend it. The plaintiffs are not the only people who can bring a claim. Private plaintiffs, state attorneys general, and other federal agencies have their own authorities, and the Bureau’s own semiannual report notes it disavows enforcement based solely on statistical correlations — a disavowal of enforcement, not a declaration that the underlying conduct is lawful, and not a limitation on anyone else’s jurisdiction.

What fewer examinations means for a 2027 program

Here is the uncomfortable arithmetic of the last eighteen months. The number of times you are examined is going down. The population of things that can be examined is not.

What actually shrank: the frequency of regulatory contact, the volume of new matters, and the breadth of products in scope. What did not shrink: the rule book your own policies are built against, the state UDAP statutes that sit beside the CFPB, the GSE selling and servicing guide requirements, the contractual standards in your lender and investor agreements, and the discovery rights of a private plaintiff who does not need a supervisory action to get your files.

There is also a second-order effect that is easy to miss. A smaller supervisory program with a larger per-matter focus means fewer low-value findings and more consequential ones. Under the new approach, examinations concentrate on clearly identified products and specific violations tied to named statutes. A single adverse action defect on a single file is now a more plausible examination outcome than a statistical review of a portfolio, because that is where the Bureau said it would look.

And the fair lending examination work has not stopped. The semiannual report is explicit that no fair lending violations were cited during the reporting period and that most supervisory work was still in the scoping stage. Scoping is not the same as finished. The work it describes for ECOA is: uncovering direct evidence of intentional discrimination on a prohibited basis, systemic failures to provide applicants with legally required notices of adverse action, and lenders’ systems for identifying risks of non-compliance with ECOA and Regulation B. For HMDA it is validating the accuracy of key data points collected and submitted.

Read that list again against the amended rule. Two of the three ECOA work-streams have almost nothing to do with disparate impact. Systemic failures to provide adverse action notices is a documentation and process test. Systems for identifying risks of non-compliance is a governance test. Both survive the rewrite intact.

Six things to have in place before January 1

1. Re-read your adverse action reasons against the new standard, and against Section 1002.9, which did not change. Section 1002.9 was never amended. Specific reasons, the reasons for the specific reasons, and the disclosure of the right to a free explanation of the file are exactly as they were. Do not let a narrow rule on disparate impact read as a general loosening of adverse action discipline.

2. Separate “what the rule permits” from “what you can defend.” The Bureau has disavowed statistical-correlation-only enforcement. Your file still has to be defensible if a private plaintiff, a state regulator, or an agency with different authority looks at it. Build the record to the stricter standard you actually face.

3. Audit your marketing, advertising, and outreach records for the discouragement change. The rule narrowed what counts as discouragement, and it narrowed it in a way that excludes location-based evidence. But your ad copy, your scripts, and your landing pages are spoken or written words or visual images, which are exactly what remains covered. Review them against the intent standard once, deliberately, and record the date.

4. Inventory any use of a protected characteristic in an SPCP, and any relationship with a for-profit SPCP originator. This is the change most likely to have a counterparty in your file without anyone telling you. Ask your business partners what their SPCP structure looks like now.

5. Document your ECOA and Regulation B risk-identification systems. The Bureau named this as examination work. If you do not have a written process for how you identify where your policies might fail — and what you do when they do — you cannot show the system the examination is looking for. This is a documentation exercise, and it is not hard. It just has to exist.

6. Build the HMDA data accuracy evidence trail while you are in there. HMDA examination work is validating the accuracy of key data points. This is a data reconciliation task, not a policy task, and it gets harder every quarter you defer it.

What to do with the files you already have

If you inherited a portfolio of declined files from before July 21, you do not need to reopen all of them. What you need is a defensible record of the standard you applied at the time and the evidence you applied it to. A dated policy version, the reason mapping in place on that date, and the adverse action notice as delivered is a reasonable answer if someone asks.

What you should not do is retroactively re-review with the new standard and change the reasons. That creates a document that contradicts the file it sits in, and it is a far worse discovery than the original decision would have been.

The larger point is about how to run CFPB compliance in 2027. Fewer examinations is not a reason to run a smaller program. It is a reason to run the program you already have more precisely, because the Bureau has told you it will focus on a small number of named things, and the cost of a defect in one of those named things is now the whole point of the visit. Fewer visits, higher consequence.

If you want a second set of eyes on where your own program sits against the amended standard, Synergy’s compliance review service works through your existing policy set and testing evidence. If you would rather walk through your own exposure on a call first, request a review through Synergy and we will tell you what we would test and what we would leave alone.

Fannie Mae Information Security: 10 Questions on the January 2027 Supplement

Fannie Mae information security compliance just got more complicated than the coverage suggests. Fannie Mae published Supplement Bulletin 26-01 on October 20, 2026, with a January 18, 2027 implementation deadline, and the change table inside it does four things at once: adds obligations, removes obligations, broadens a trigger you thought you understood, and adds a clause under which silence counts as acceptance. Most lenders will read the MFA and encryption lines and stop there. Here are the ten questions lenders are asking, with direct answers and the action each one calls for. Several items were still moving as this edition went to press; where an answer depends on a pending confirmation, we flag it.

Compliance guidance for discussion, not legal advice. Requirements vary by lender, product, and state. Consult counsel on your facts.

What Changed on January 18, 2027

1. What actually changes on January 18, 2027?

Four things, and only one is a new control. Bulletin 26-01 adds a new Section 4 on super intelligence governance, names NIST Framework 800-53 as an example alignment standard, adds an insurance policy requirement, and replaces the threat assessment with a business impact analysis. It removes obligations: the reference letting a Single-Family lender deliver the annual attestation through Form 582, contingencies for fourth-party vendors, sharing lessons learned with management, and the annual threat risk assessment. It broadens the recovery trigger in your business continuity plan, expands what you must report after a cyber incident, and replaces “vendor” with “third party” throughout. And it adds a clause most people will miss: a Company is deemed to accept the amendment if it does not reject it in the applicable period. Read the change table before you build the implementation plan.

2. Does the artificial-intelligence-to-super-intelligence rename change anything?


Two different things, and the difference is where the work is. The word changed, and it changed nothing: the Supplement’s FAQ says the terminology update “is not intended to alter the scope, meaning, or applicability of the Supplement’s requirements.” The section is new, and it is not nothing. Bulletin 26-01 adds Section 4, Super Intelligence, with governance requirements for Companies using SI in connection with activities under or relating to their Underlying Agreements — an SI Governance Program, SI supply chain management, and disclosure to Fannie Mae. Fannie places the “scope, meaning and applicability remain unchanged” line directly above a brand-new SI section. Both are true; only one is about the rename.

For most lenders the SI Governance Program documents work already underway: LL-2026-04 already required AI/ML policies, Supplement compliance, no-less-protective governance of vendor and subcontractor AI use, and disclosure to Fannie Mae on request, and Multifamily lenders have the parallel LL-2026-05. Our AI governance Q&A covers that. Build the SI program as a standalone document anyway, because Section 4 makes it a Supplement requirement in its own right.

3. Did the section numbers change?


Yes, and the old numbers now point somewhere else. Cybersecurity Incident Management moved from Section 4 to Section 5; Actions by Fannie Mae is now 5.1; Additional Company Requirements, which carries the loss allocation, is now 5.3. The Business Continuity Plan moved from Section 5.1 to 6.1, and Section 4 is now Super Intelligence. Two consequences. If your policies, vendor contracts or audit scope cite “Supplement Section 4” for incident reporting, they now point at the AI governance section. And any mapping you built against the current Supplement has to be renumbered before the attestation, because the officer signing it is attesting to the new edition.

The Two Reporting Clocks


4. When does the 36-hour clock start?


Earlier than most teams plan for. The trigger is identification of a Cybersecurity Incident or the reasonable conclusion that one may have occurred — and “we think something may have happened and we are still investigating” is inside the clock. That makes this a compliance obligation rather than a security one, and why the incident response plan needs severity assessment, call trees and evidence collection. A vendor event starts your clock too, because the obligation reaches the Company or its third parties.

Two things to note. The 36-hour standard did not change, but new Section 5 splits reporting into initial notification inside the 36 hours and prompt notification of supplemental details afterward — you are not expected to hold the full forensic picture at hour 36, only to have made the call and kept reporting current. And 6.2 removed the requirement that your business continuity plan share lessons learned with management and track them to resolution. Annual testing survives, and technology continuity planning is now disaster recovery planning.

5. What has to be in the first report?


To the Fannie Mae Privacy Office at privacy_office@fanniemae.com, or by whatever other means Fannie Mae requests, then promptly as requested. The initial notification carries the identification date, mitigation and containment status with the actions taken, and whether business operations were suspended. Supplemental details follow promptly: indicators of compromise, attack methodology, third-party providers, law enforcement, the investigation lead, and extent of exposure. Lost, stolen or incorrectly routed physical information carries the same 36-hour notice.

What is new in Section 5.3 is the list of what you report. Reporting now includes all data attributes of impacted individuals, and the notification letter you sent to impacted individuals, with a copy to Fannie Mae. Your individual breach notice is now a Fannie Mae deliverable, not only a state-law obligation, and if a vendor handles consumer breach notices for you, that vendor sits in your reporting chain, not just your security chain. Plan that in advance, because none of it is available on day one unless somebody is assigned to assemble it.

6. Do I have to report a disruption that is not a cyber incident?


Yes, and on a shorter clock than most assume. The Supplement has always required notice of disruptions, including technology-related ones, that may impair business conducted with Fannie Mae or your ability to perform your obligations — even where no Cybersecurity Incident occurred. What Bulletin 26-01 adds is one word: Section 3.6 clarifies that notice of any disruption must be “prompt.” That is a shorter fuse than the 36-hour cyber window, so your disruption escalation path cannot be slower than your incident path.

The business continuity trigger widened too. Section 6.1 now requires your plan to let you recover from a disruption of your relationship with a third party that is material to business conducted with Fannie Mae — broader than the old “expiration or termination of a material contract” language, so a degrading vendor relationship is in scope. Two items came off the list: contingencies for fourth-party vendors, and recovering business operations for third-party vendor issues, now handled by the parallel Section 3.14 requirements. That is relief you have to take deliberately — the work moves into the vendor contract.

Attestation, Liability and Third Parties


7. Can I still deliver the annual attestation through Form 582?

No, and this is the change most likely to catch you. Bulletin 26-01 removed from Section 3.1 the reference to a Single-Family lender providing the attestation through the annual Form 582 process. The requirement itself is not removed — you still owe a written attestation, executed by a duly authorized corporate officer, that your Information Security Program and associated plans, policies, standards and procedures meet the Supplement. What is removed is the off-ramp. If you have been treating the attestation as a line in a form you file anyway, treat it as a direct deliverable, and confirm the submission channel with your Fannie Mae contact: we could not verify whether Form 582 remains an optional route, and the Form 582 FAQ does not address it.

Two related changes sit in the same edition. NIST Framework 800-53 is now named as an example alignment standard alongside ISO 27001. And Section 3.3 changed the independent security assessment requirement from a “qualified independent auditor” to an “independent third party security professional,” and confirmed that the Company must commission it. That requirement has history — the September 2025 bulletin removed the prohibition on an affiliated assessor — so the test is not independence in the corporate-governance sense, it is that a qualified third party did the work and you commissioned it.

8. Who is responsible for the losses?

You are, and the new edition makes the reporting match. The September 2025 Supplement Bulletin added the Company as the responsible party for all Losses arising from a Cybersecurity Incident — Section 4.3 in the current edition, Section 5.3 in the new one. That is a change in who carries the loss, not just who files the report, and the reason to know within hours which vendors hold Fannie Mae Confidential Information and which were in the blast radius. The same section now expands what you report: all data attributes of impacted individuals, and the notification letter you sent to impacted individuals, with a copy to Fannie Mae. Put the two together and the exposure is clear — you own the loss, you own the forensic detail, and you hand Fannie Mae the letter you wrote to the consumer.

9. How far down the vendor chain do these obligations reach?


Further than most vendor contracts support, and the vocabulary just changed. Bulletin 26-01 replaced the term “vendor” with “third party” throughout the Supplement and updated the Company’s third-party compliance requirements, so every vendor agreement and checklist that quotes the Supplement is now quoting the wrong word.

The substance holds. A Company using service providers to store, process, access or transmit Fannie Mae Confidential Information must require substantially similar information security and business continuity requirements, under Section 3.14, Supply Chain Risk Management. Selling Guide A4-1-01 runs the same line: you and any subservicers, third-party originators, outsourcing firms and third-party vendors you use must maintain disaster recovery and business continuity procedures. And if your fulfillment or back-office partner, document custodian or core origination vendor has never been asked to evidence a security program, a business continuity plan, a tested incident response plan and a 36-hour notification commitment back to you, that gap is yours. Document custodians carry their own line under the Requirements for Document Custodians: report to Fannie and any affected lender within 10 business days after a single loss in their custody division exceeding $100,000.

The Most Common Failure


10. What is the most common Supplement compliance failure you see?

Treating the Supplement as a controls update instead of a change to what you owe. Bulletin 26-01 is easy to read that way, because the most visible lines are MFA requirements, updated encryption language and a new section number. A control map shows you all three. It does not show you the removed Form 582 route, the dropped fourth-party contingency requirement, the widened third-party recovery trigger, the requirement to hand Fannie Mae your consumer breach notice letter, or the clause that deems you to have accepted the amendment if you do nothing. Four of those five are invisible on a control spreadsheet, and three carry deadlines.

The fix is not a longer binder. It is a document set that matches the new edition section by section, a named person who owns the 36-hour call, a notice requirement in every third-party contract, and a decision on the record about whether you are accepting the amendment. Those are four documents and one meeting. The lenders who get an unpleasant conversation with Fannie Mae will not be the ones with the longest security policy. They will be the ones who never decided who owns the 36-hour call and never noticed the 36-hour deadline was the easy part.

Want help diffing Bulletin 26-01 against your information security program, standing up the attestation, or tightening the notice language in your vendor contracts? Synergy supports mortgage banks and credit unions that have to evidence this on the record. Book a 30-minute call and we will send our one-page Supplement readiness checklist.

Adverse Action Notices: What the CFPB Will Actually Examine Now

Thursday, 9:40 a.m. A fair lending review pulls thirty declined applications from last month. Every adverse action notice is on time. Every reason is specific. Every disclosure is signed. On file 19, the notice lists two reasons — and neither one is why the file was declined. The loan officer’s notes say something else entirely.

Nobody approved that step. Nobody owns it. The reason codes on those adverse action notices are not in the QC checklist, the mapping from policy screen to decline reason is not in the exception log, and nobody has reconciled a reason to what drove the decision since the models changed.

This is the least dramatic version of what just happened to mortgage compliance.

What the CFPB Published on October 1


On October 1, 2026, the CFPB released its Fair Lending Report for 2025 to Congress. The enforcement numbers look like a retreat. The Bureau closed 76% of its supervisory actions — nearly 1,500 matters — roughly 12% of them fair-lending related. It initiated no new fair lending supervisory activity, cited no institution for a fair lending violation, and made no ECOA referrals to the Department of Justice. Across the eleven agencies with ECOA enforcement authority, 140 institutions were cited for ECOA or Regulation B violations, and no public ECOA enforcement action was brought in 2025.

Here is the part that should change your review schedule. The same report says ECOA examination work now focuses on three things: direct evidence of intentional discrimination on a prohibited basis, systemic failures to provide applicants with legally required notices of adverse actions, and the systems a lender uses to identify ECOA and Regulation B risk. HMDA exam work focuses on validating the accuracy of key data points.

Read that list again. Two of the three ECOA exam priorities are about the file in front of you, not the portfolio behind you. The statistical screen you scaled back in July is not what a 2026 exam is testing. The notice is.

Then on October 14, the OCC released Guaranteeing Fair Banking for All Americans, its review of debanking practices at the nine largest national banks it supervises — JPMorgan Chase, Bank of America, Citibank, Wells Fargo, U.S. Bank, Capital One, PNC, TD and BMO. The review covered September 2020 through September 2025, drawing on nearly 100,000 complaints and about 50,000 documents totaling roughly 420,000 pages. The OCC found broad reputation risk policies lacking clear standards and objective, risk-based criteria, plus industry-specific restrictions on oil and gas, coal, firearms, detention facilities, tobacco and digital assets.

Be precise about what that report does and does not say. It identifies no violation of law and announces no enforcement action. It did remove reputation risk from the OCC’s own supervisory framework, issue Bulletin 2025-22 allowing debanking to be considered in licensing and CRA exams, and signal further rulemaking. Hours later, the CFPB announced it will fully investigate whether the institutions in the report and others complied with federal consumer financial laws — naming ECOA, FCRA “particularly the adverse action notice requirements,” and UDAAP.

Four Risks That Replaced the Statistical Screen


Stripped down, a fair lending file is one of four things, and they are nowhere near equal in risk. The tiers are our suggested placements for your own scoring — no examiner grades declination files for you, so treat them as a starting position, not an official classification.

The inaccurate reason. The notice is timely, specific, and still wrong, because the reason the borrower was given is not the reason the file was decided. The CFPB has named systemic failures in adverse action notices as an exam priority, so a notice that cannot be tied to the decision is a finding waiting for a file. Our suggested tier: High.

The reputation-risk screen. A decline that passed through a policy screen — an industry, a brand, an ESG or reputation review, a manual escalation — and left no trace on the notice, because the screen is not a “credit reason.” The risk is what the OCC documented: criteria broad enough to apply inconsistently, and outcomes nobody can reconstruct later. Our suggested tier: High.

The stale notice regime. Templates, timing and disclosures that were correct in 2024 and have not been re-read since. Section 1002.9 was not amended by the April rule, so the 30-day notice requirement and the specific-reasons standard still apply exactly as they did. The risk is that the file reads like a rulebook from a year ago. Our suggested tier: Moderate.

The data that does not reconcile. HMDA data points that do not tie back to the source document, or a reason code in the loan system but not on the notice. The exam focus the CFPB named for HMDA is the accuracy of the key data points. Our suggested tier: Moderate.

The line worth memorizing: a statistical test tells you what your portfolio looks like. An adverse action notice tells the examiner what your lender did to one applicant.

What the Rules Actually Say Now


Last verified against primary sources October 21, 2026. This is a moving target — state law changes and the regulators are still writing. Re-verify before you rely on any of it.

What the April rule changed, and what it did not


Start with what did not change, because the volume of material saying otherwise is the problem. The CFPB’s final rule amending Regulation B was published April 22, 2026 and took effect July 21, 2026 (91 FR 21620, Docket No. CFPB-2025-0039). It removed the effects test and states ECOA does not authorize disparate impact liability. It did not amend 12 CFR 1002.9: the duty to give the specific principal reasons for adverse action within 30 days of a completed application is untouched, as is FCRA’s adverse action requirement and the credit score key factor disclosure. Disparate treatment survives in full under ECOA.

And the rule amended Regulation B only. Disparate impact remains available under the Fair Housing Act, which the CFPB cannot amend, under HUD’s FHA disparate impact standard, whose rulemaking is still open, and under state regimes — New Jersey, New York, and Illinois from January 1, 2027 — plus private litigation. We covered that in why state fair lending rules now drive your testing and in what the ECOA rule eliminated and what lenders must fix in fair lending QC. The short version has not changed.

The scope of what the CFPB can actually reach


One scope point, and it matters. ECOA reaches credit, not deposit accounts, and does not list political beliefs as a prohibited basis. The OCC placed state-law compliance outside its scope. So the live theories are the three the CFPB named: ECOA, FCRA — especially the adverse action notice requirements — and UDAAP. Debanking is the context. The document you can be held to is the notice.

Watch three things. HUD’s FHA disparate impact rulemaking. State adoption of new testing expectations, and the multi-state letter opposing the federal rule. And further OCC rulemaking on reputation risk and pretextual account terminations.

The risk nobody has absorbed


The risk most operators have not absorbed is the mapping. The CFPB now describes an exam that looks for a lender’s systems for identifying ECOA risk and for systemic failures in adverse action notices. Accurate reasons with no evidence that the reasons came from the decision is half of what that exam needs. The Massachusetts Attorney General’s 2025 Earnest settlement is the most detailed published template of what a state regulator expects: interpretable models, every decline reason validated, principal reasons stated specifically and accurately. It binds one lender, not you — but it is a template you can be examined against.

The Numbers Vendors Want You to Believe


Before the controls, a word about the arithmetic, because these numbers will reach a board deck whether you are ready or not. The headline number in fair lending software right now is relief: disparate impact testing is no longer a federal requirement, the effects test is gone, and the statistical review that ate a quarter of your compliance calendar can be scaled back. For ECOA that is largely true, and we wrote about it.

The second number is certainty, and that is where the risk is. Some vendors still sell disparate impact testing as the requirement it no longer is. Some sell “compliant with the 2026 Reg B” as if the April rule resolved fair lending exposure. Others quote guidance the CFPB has withdrawn — 67 documents, seven fair-lending related, in May 2026 — while vendor pages kept citing a “CFPB Circular 2026-03” that does not exist.

  • Which regime are you actually answering to? A vendor’s “compliant” claim is a claim about ECOA. Your book is also covered by the Fair Housing Act, HUD’s FHA standard, and the state regimes where you operate. Ask which regimes it maps to.
  • What is the denominator? Coverage of a statistical test is not coverage of the notice. A product that scans your portfolio and reports a pass has not tested one declination notice.
  • What is the measurement window? Ask when the vendor last reviewed its regulatory content against the April 22 rule and the July 21 effective date. Where guidance is withdrawn quarterly, content with a 2024 review date is a liability.

Keep the ROI column separate from the evidence column. A lender with both a credible efficiency claim and a defensible file walks into a review with an answer. A lender with only the first one is negotiating one.

Six Controls to Put Around Your Adverse Action Process


None of this requires a data scientist. All of it is something an examiner can ask for on the record.

  1. Test the notice against the file, not the template. Pull the notice and the credit file together. Confirm the reasons on the notice are reasons that appear in the file, and that nothing else drove the outcome.
  2. Map every decline reason to what actually drove the decision. If a model or a policy screen is in the path, document the mapping and validate it the way the Massachusetts settlement describes: interpretable models, every reason validated, no reason invented to fill a menu.
  3. Inventory every screen that can end a file. Reputation risk, brand, ESG, industry, manual escalation. For each, write down the objective criteria, who applies it, and where it shows up in the file. A screen that can decline a file and never appears on the notice is the gap the OCC documented.
  4. Stop letting reason codes be a dropdown. The Earnest assurance alleged that staff picked an available reason when the real one was not on the menu, and that auto-generated reasons sometimes produced nothing specific. Both are cheap to prevent, expensive to catch.
  5. Reconcile your HMDA data points to source. The exam focus is accuracy. Sample the fields, trace them to source, and fix the definition, not the file.
  6. Date and version your fair lending policy. Record what changed, when, on whose authority, and re-verify the citations before each exam cycle. A 2024 review date is a finding waiting to happen.


Where to Start on Monday


You do not need a program. You need a week. Pull thirty declination files closed in the last 90 days. For each, produce the notice, the credit file, the reason-code mapping, the source of the decision, the policy behind any screen, and the HMDA data points. Any gap you find now is a roadmap. The same gap, found by an examiner, is a finding — and if you would rather someone else pull the sample, our fair lending compliance assessment runs that review.

Adverse action notices are the cheapest artifact in a mortgage file and the most expensive one to get wrong. The CFPB has told you what it is looking at. Now it is whether your file can answer it.

Compliance guidance for discussion, not legal advice. Requirements vary by lender, product, and state, and state adoption is still moving. Consult counsel on your facts.

Synergy supports mortgage banks and credit unions that have to answer for fair lending in a file, not a slide deck — adverse action notice review, declination file testing, and policy work that holds up on the record. Book a 30-minute call and we will send our one-page declination review checklist for last quarter’s files.

Mortgage AI copilots are now inside the loan file, and the GSEs expect you to govern them.

It is 4:47 on a Wednesday. A processor has a file stuck on a condition and a closing that is not going to happen tomorrow. She types one question into a box on her screen — does this paystub plus bonus clear the income condition, or do I need a written addendum? — and gets a grounded, guideline-referenced answer in nine seconds. She clears the condition, the file moves, the closing gets made.

Nobody approved that software. Nobody risk-tiered it. It is not on the vendor list, not in the policy manual, and not in the AI inventory the GSEs now expect you to produce on request.

This is the least dramatic version of what just happened to mortgage origination.

Mortgage AI Copilots Launched This Week


This is the week mortgage AI copilots stopped being a demo and started being a line item. On October 12, 2026, nCino introduced a Loan Officer Productivity Suite — six AI tools folded into its mortgage point of sale, including a conversational advisor for pipeline and GSE questions, an eligibility pre-check, and a connector that lets a lender plug its own AI agents straight into the file. The same day, Blend announced a tenth lender had signed on to Blend Autopilot, its agentic layer for home lending, now running more than 3,500 new loans a week.

Four days earlier, on October 8, origination platform Vesta closed a $30 million Series B to expand AI agents that work inside the lender’s system of record — reviewing documents, clearing conditions, checking closing packages, making basic underwriting calls. Vesta says roughly 40% of tasks on its platform now route to agents and automated workflows, with every action logged for audit.

Four Copilot Types, Four Risk Levels


Stripped down, mortgage AI copilots fall into four buckets. The tiers below are our suggested placements, using the CSBS worksheet’s Low / Moderate / High definitions. The worksheet tiers individual use cases, not these four categories — treat them as a starting position for your own scoring, not an official classification.

  • Guideline Q&A. Answers a plain-language question about qualifying income, gift funds, or documentation, grounded in GSE and agency guidelines. The risk is hallucinated or stale guidance — a loan officer acting on a confident wrong answer. Our suggested tier: Low–Moderate.
  • AUS interpretation. Translates automated underwriting findings into a plain-language summary and a list of suggested conditions. The risk is a misread finding becoming a live file condition nobody independently verified. Our suggested tier: Moderate–High.
  • File action. Creates tasks, sends reminders, chases conditions, and populates the application, typically with originator approval. The risk is unlogged or unapproved action on a live file, with no evidence of who did what. Our suggested tier: Moderate–High.
  • Agentic work. Reads the file, decides what happens next, and acts — document review, income review, pre-underwriting, closing QC. The risk is delegated judgment on loans headed to the GSEs, behind a weak audit trail. Our suggested tier: High.

The line worth memorizing: a tool that drafts an email is a productivity story. A tool that clears a condition on a loan you intend to sell is a governance event — and it is happening inside rules that have been enforceable for months.

Fannie Mae, Freddie Mac and CSBS AI Governance Requirements


Last verified against primary sources October 14, 2026. This is a moving target — re-verify before you rely on any of it.

If your mortgage AI copilots are clearing conditions, drafting disclosures, or influencing eligibility, all three frameworks reach them — and the rest of this section is the part your examiner will be looking for.

Fannie Mae has been live since August 6. Lender Letter LL-2026-04 requires written policies covering the development, use, and maintenance of any AI or ML system in origination or servicing — reviewed at least annually, incorporating trustworthy and ethical AI characteristics, aligned to your stated risk tolerance. It draws no line between tools you built and tools a vendor built, requires vendor AI risk to be governed no less protectively than your own, and obligates you to disclose on request the types of AI in use, the purpose and manner of use, and your safeguards.

One scope point: LL-2026-04 binds Fannie Mae sellers and servicers. Sell only to Freddie Mac and it does not bind you — but Freddie’s own requirement does, and it is the harder one.

Freddie Mac got there first and went further. Guide Bulletin 2025-16, issued December 3, 2025, rewrote Seller/Servicer Guide Section 1302.8 with a framework effective March 3, 2026: senior management approval, documented accountability, monitoring for performance degradation and bias, threat assessment for data poisoning and adversarial inputs, and regular internal and external audits against standards such as NIST 800-53 and ISO 27001. Freddie also carries an express indemnification tied to AI use. Sell to both and you satisfy the harder elements of each.

Your state regulator published the roadmap on September 16. The CSBS Artificial Intelligence Supervisory Framework is discretionary and creates no new legal obligation — but it tells state examiners what to ask: does the firm use AI, where, how does it touch customers and shape decisions, did it come from vendors, is AI embedded in platforms you already run, do you tier your use cases, and does sensitive information pass through it. The optional Risk Tiering Worksheet sorts use cases Low, Moderate, or High on consumer impact, human oversight, harm potential, and data sensitivity, with suggested controls scaling by tier — from inventory and acceptable-use policies at the low end toward independent model validation and dedicated incident response procedures at the highest.

The revised interagency model risk management guidance issued April 17, 2026 (OCC Bulletin 2026-13, SR 26-2, FDIC FIL-15-2026) expressly carved generative and agentic AI out of its scope, calling it “novel and rapidly evolving.” So the federal model risk rule will not hand you a copilot checklist — Fannie, Freddie, and your state examiner will. Note that this guidance is supervisory, not enforceable, and it applies to banks rather than nonbank mortgage lenders.

Expect that to change: the same announcement states the OCC, Federal Reserve, and FDIC plan to issue a request for information on model risk management, with particular attention to banks’ use of AI. That RFI belongs on your watch list.

The consumer-protection question is the one most lenders have not absorbed. The CSBS Consumer Protection supplement points examiners toward whether data used by an AI system — including proxies and alternative data — could produce inaccurate or unfair consumer outcomes, and whether meaningful human review is actually applied. Copilots trained on historical lender data inherit historical lender patterns, and the loan officer’s name on the file does not cure that. Our fair lending compliance assessment is where lenders should start when AI already touches credit decisions.

The ROI Numbers Vendors Want You to Believe


Before the controls, a word about the business case for mortgage AI copilots, because these numbers will reach a board deck whether you are ready or not. According to Blend, lenders using Autopilot report 10% to 15% higher pull-through, cycle times shortened by two to four days, and an average of 4.5 hours of fulfillment work automated per loan. Blend also notes those results compare each lender against its own pre-Autopilot performance — self-reported against a customer-supplied baseline, not an independent audit.

The Vesta numbers show why that distinction matters. Vesta cites Pennymac’s 25% cut in origination operating costs and roughly 50% loan officer efficiency gain — but both came from replacing its legacy LOS, before the lender activated a single AI agent. The efficiency belonged to the platform migration, not the AI. Ask every vendor for the denominator, the measurement window, and the baseline, and keep the ROI column separate from the control evidence. A lender that can produce both a credible efficiency claim and a defensible governance file leaves the exam with a “compliant.” A lender with only the first one is negotiating a finding.

Six Controls to Put Around Mortgage AI Copilots


None requires a data scientist. All are things an examiner can ask for on the record.

  • Build the inventory from the platform outward. Start with the LOS you run today and the vendors you already pay. Copilots rarely arrive as new line items — they arrive as features. Blend now ships an MCP server so lenders can bring their own agents, and nCino folded six capabilities into a POS your staff already logs into. An inventory that only lists products marketed as “AI” will be wrong inside a quarter.
  • Tier every use case. Use the CSBS worksheet as your scoring instrument and document the tier, the business owner, and the human checkpoint for each one.
  • Make the vendor show receipts — then put it in the contract. Get past the demo. Ask how the system was trained and validated, whether non-public personal information trains or improves the model, what is retained, how performance and bias are monitored, how often the rules change, and what documentation it will supply for your GSE or state disclosure requests. Then re-paper the contract around the answers: model change notice, monitoring, audit rights, incident notification, and contingency planning for what happens if the model becomes unavailable or starts producing unreliable output.
  • Make every action loggable and reversible. The stronger copilot vendors already record reasoning step by step, route anything needing judgment to a human, and log every action. Your standard should require the same — permission boundaries, human checkpoints, and the ability to halt an action are what the CSBS Examiner Work Program looks for in agentic systems.
  • Write the acceptable-use policy, including shadow AI. Most staff now have a general-purpose AI assistant in the same browser tab as the borrower file. The Framework expects rules on approved tools, what data may be entered, how output is reviewed, and when AI material may be used externally — plus the ability to identify unauthorized use. If your policy says nothing, that gap is documented and visible.


Where to Start on Monday


You do not need a program. You need a first week, and a lot of lenders already have mortgage AI copilots in the stack without an owner. Pull every tool your team touches on a live file — including the browser assistant. Tier them, name an owner for each, and ask the top three vendors the questions in control three, in writing. Then run the mock exam CSBS describes: could you produce, on request, an AI inventory, your policies, risk assessments, vendor due diligence, testing results, training materials, and evidence of management oversight? Any gap you find now is a roadmap. The same gap found by an examiner is a finding — and if you want a second set of eyes before the next exam cycle, our compliance services team runs exactly that review.

Compliance guidance for discussion, not legal advice. Requirements vary by lender, product, and state, and state adoption of the CSBS framework varies by agency. Consult counsel on your facts.

Synergy supports lenders navigating AI-enabled mortgage operations with practical, evidence-based compliance review — AI inventory development, vendor due diligence, and control design built around the frameworks above. Book a 30-minute call and we will send our one-page AI inventory template so you can run it against your vendor list this week.

What is Mortgage Processing?

Mortgage processing is the work that turns a signed loan application into a funded loan — and it is where most home purchases are won, lost, or delayed.

The borrower signs a Form 1003. The loan officer promises a closing date. And then, for the next two to four weeks, the file lives with the processor.

That middle stretch gets called “the boring part.” It is the part that decides whether a family gets keys on the contract date or signs a six-week extension. Here is what it actually involves, who owns each piece, and where it breaks down.

What Mortgage Processing Actually Covers


Mortgage processing starts the moment a loan officer hands off a file, and it ends when the loan is funded and delivered. In between, a processor does three jobs that are easy to blur together: they verify the file, they chase what is missing, and they keep the timeline.

Concretely, that means pulling and ordering income and asset documentation, running the credit report, ordering the appraisal, title work, and flood certification, entering data into the loan origination system, and assembling the package an underwriter can actually read. Once the loan is approved, processing continues: producing the closing package, coordinating with the title or escrow company, confirming every condition is cleared, and getting the file to clear to close.

It is a documentation role with a deadline problem attached. Almost everything a processor does exists because a rule, a guideline, or a borrower request demands it — and every one of those demands has a clock attached.

For a lender, mortgage processing is the largest controllable cost in the file and the biggest controllable driver of cycle time. A file that stalls is a file that costs money to carry and a borrower who may walk away. That is why shops measure turnaround at every stage rather than only close rate, and why the first three weeks of processing tend to decide whether a loan is profitable before anyone opens the margin discussion.

The Stages of Mortgage Processing, Step by Step


Every file moves through the same sequence, whether the loan is a first mortgage or a forty-file production week.

  • Application and Loan Estimate. The borrower provides six pieces of information — name, income, Social Security number, property address, estimated value, and loan amount. The lender owes a Loan Estimate within three business days of receiving them. On a refinance, the interest rate is usually locked at this point, which starts the real countdown.
  • Document collection. Pay stubs, W-2s, recent tax returns, bank statements, and asset statements come in and get verified against what the borrower claimed. This is the stage where most borrower friction lives.
  • Ordering third-party work. Appraisal, title commitment or preliminary report, flood certification, and verification of employment and income. The processor owns chasing these vendors, which is a full-time job when a valuation company is backed up.
  • Automated underwriting. The file runs through the AUS, which returns findings or a refer. Most files are decided here; some come back with conditions or get pulled for a human underwriter.
  • Underwriting decision. An underwriter verifies the file complies with the product guidelines and issues an approval, a conditional approval with conditions to clear, or a denial. Fannie Mae’s Selling Guide, for example, requires the final signed Form 1003 in the file, reflecting the income, assets, debts, and final terms used in underwriting.
  • Clear to close. Every condition — borrower and underwriter — is satisfied and documented. The file is final.
  • Closing and funding. The title or escrow company conducts the closing, the borrower signs the note and security instrument, the lender funds, and the file ships to the investor.

Who Does What on a Processing File

Mortgage processing goes wrong when nobody owns a step. Five roles touch every file, and their boundaries are worth being precise about.

  • Loan officer. Originates the relationship and the loan, and is the processor’s customer. The processor verifies what the borrower told the loan officer; they do not re-sell the loan or make credit decisions.
  • Processor. Builds, verifies, and tracks the file. Sits between the loan officer and the underwriter, and is usually the borrower’s main point of contact once the loan officer steps back. In many shops the processor is also the only person looking at the file every day.
  • Underwriter. Makes the credit decision and writes the conditions. Typically does not gather documents directly.
  • Closer. Picks up after clear to close — orders and reviews the closing package, coordinates the settlement agent, and manages funding and investor delivery.
  • Title and escrow company. The settlement agent outside the lender. They run the closing, hold the money, and record the lien. They are an outside party with their own deadlines, and they are a very common source of delay.


Our mortgage loan fulfillment services cover the processing, underwriting, and closing stages together, because in practice the hand-offs between them are where errors get made.

The Disclosures and Deadlines That Drive the File


Mortgage processing is governed by a small number of hard dates under the TILA-RESPA integrated disclosure rules. Get one wrong and the closing cannot happen on time, no matter how clean the file is.

  • Loan Estimate: three business days. Due no later than the third business day after the lender receives the application, and it must also be provided no later than seven business days before consummation.
  • Revised Loan Estimate: four business days. If a changed circumstance allows a revision, the revised disclosure generally must reach the borrower no later than four business days before consummation.
  • Ten business days to express intent. If the borrower does not indicate they want to proceed within ten business days of receiving the Loan Estimate, the lender may treat the application as abandoned or use a revised estimate.
  • Closing Disclosure: three business days before closing. The borrower must receive it at least three business days before consummation. Here is the detail most summaries get wrong: on a purchase, only the primary obligor is required to receive it — but on a refinance, every borrower who signs the note and every individual with an ownership interest who has a right to rescind must receive it. That is a different list, and a common source of last-minute holds.
  • Right to cancel: three business days, refinances only. A purchase-money mortgage carries no right of rescission. This is the single most misunderstood point in consumer mortgage compliance.


One more rule that catches people: once the Closing Disclosure is issued, no further Loan Estimates can be issued. The file is in the final disclosure lane. The CFPB’s closing guidance lays out what a borrower should be reviewing in that window.

Where Mortgage Processing Files Stall


Cycle time is a function of what is missing, not how hard anyone is working. The recurring culprits:

  • Income and asset documentation that does not match what the borrower stated on the application — a changed job, a new account, a gap that needs a letter. Every mismatch becomes a condition.
  • Third-party vendor queues on appraisals, title, and flood determinations. Processing cannot fix a valuation company that is three weeks out.
  • Borrower responsiveness. Most conditions go to a person juggling a move, a new job, and a closing date. Chasing is a large share of a processor’s actual workload.
  • Condition drift between departments. A closer working from a stale version of the file is a classic source of rework, which is exactly what our all-validation review process is built to catch.
  • Discrepancy problems that surface late — a final CD figure that moved too far from the Loan Estimate, or a funding figure that will not reconcile.


The last one is worth dwelling on, because mortgage processing failures are rarely the dramatic kind. A file does not blow up. It just ages, and then it closes late, and then a quality review finds something that should have been caught before funding.

What Changes Next on the Processing Desk


Mortgage processing is in the middle of a tooling shift. Vendors now sell agents that read the file, clear conditions, and work inside the loan origination system alongside the humans — the same shift described in our piece on mortgage AI copilots. Vendors report meaningful efficiency gains from these platforms, though the results are measured against each lender’s own baseline rather than an independent audit, which is worth keeping in mind.

The durable part of mortgage processing does not change. Verification, completeness, condition tracking, and a documentation trail that holds up to review are what a lender is actually buying when it buys processing. Those are judgment calls made at volume, and no amount of tooling removes the judgment.

General information for education, not legal advice. Disclosure and rescission rules vary by transaction type and state. Consult counsel on specific facts.

Synergy supports lenders with practical, evidence-based review of the files that mortgage processing produces — condition tracking, disclosure timing, and closing package accuracy across purchase, refinance, and government-loan files. If your team wants a second set of eyes before the next QC cycle, book a 30-minute call.

Mortgage AI Governance Q&A: 10 Questions Compliance Teams Are Asking in Q4 2026

The most-examined answer on your mortgage AI governance program is no longer a policy promise. It is the answer to “show me your AI inventory, your vendor AI questionnaires, your risk tiering, and your monitoring evidence.” Here are the ten questions compliance officers, QC managers, and risk leaders are asking as Q4 2026 begins, with direct answers and the action each one calls for. Several items were still moving as this edition went to press; where an answer is dependent on a pending development, we say so and flag it to verify.

Building the AI Inventory

  1. Where do we start with our AI inventory?
    Start with the inventory. It is the foundation of every other AI governance control, and the thing both GSEs and the CSBS AI Supervisory Framework ask for first. The inventory should list every AI or machine-learning use case, internal or vendor-supplied, that touches the mortgage lifecycle. For each entry, capture the business function, the system owner, the vendor (if any), what data the system consumes, whether it touches nonpublic personal information, what decision or output it influences, and whether a human reviews the output before it affects a borrower. If you cannot produce this list in a day, that gap is your first finding.
  2. What counts as “AI” for purposes of the GSE requirements?
    Both Fannie Mae and Freddie Mac define AI broadly, and “use” is broader than most lenders assume. The categories that count include automated underwriting systems, automated valuation models, fraud detection models, document classification and data extraction tools, income and asset calculation engines, chatbots answering borrower questions, servicing call-summary tools, marketing optimization models, lead scoring, and any QC platform that flags defects with machine learning. Rule-based decision engines that do not learn from data and standard business intelligence dashboards generally do not count. Verify the current text of LL-2026-04 and Guide Section 1302.8 for the precise scope.
  3. How do we tier our AI use cases by risk?
    Not every use case deserves the same level of scrutiny. A tool that summarizes internal policy documents is low risk. A tool that calculates qualifying income, influences pricing, decides which files get escalated, or drafts adverse action language is high risk because it can affect credit decisions, fair lending outcomes, or required disclosures. The CSBS AI Use Case Risk Tiering Worksheet is a reasonable starting template. The tier assignment itself is the documentation — record it, the reasoning, and the date. Tiering is not a one-time exercise. Reassess whenever a use case changes materially, when an examiner challenges a classification, or at least annually.

Vendor Oversight


4. What should be in our vendor AI questionnaire?
Most lender vendor questionnaires are not built to surface AI risk. The questions to add: Does the product use AI or machine learning anywhere in its workflow? For what specific function? How is the model tested, validated, and monitored? What data trains the model, and where is that data processed? Does the product use generative AI, agentic AI, or any large language model component? How are errors reported, and on what timeline? What does the vendor do before a material model change ships? Will the vendor cooperate with GSE or state regulator information requests? Fannie Mae’s “no less protective” standard means the vendor’s AI is your AI for governance purposes, so the questionnaire matters.

  1. Do the GSE AI requirements apply if we only use AI through vendors?
    Yes. Freddie Mac Guide Section 1302.8 applies to your use of AI or machine learning in origination or servicing, regardless of who built it. Fannie Mae LL-2026-04 requires you to manage vendor and subcontractor AI/ML risk with governance no less protective than its own requirements. Both GSEs can request details about your AI use on demand, including your vendor AI. If a vendor’s tool touches loans you sell to or service for the GSEs, it belongs in your inventory, your tiering, and your oversight program.
  2. How often should we monitor high-tier AI use cases?
    Define what “monitoring” means for each high-tier use case, and the cadence should match the risk. For document extraction and income calculation tools, the cleanest control is QC sampling: compare AI outputs to human-verified results on a defined sample each month and trend the error rate. For models that influence pricing or adverse action, the standard is higher — pre-deployment testing, segment-level performance review at least annually, and ongoing monitoring for drift, accuracy, and bias proxies. The September 2026 AARMR, MBA, and BCG survey found that only 58% of lenders monitor AI tools after deployment; the lenders who do will be the ones that hold up in a GSE request or state exam.

Examiner and GSE Readiness


7. What does the CSBS AI Supervisory Framework actually ask for?
CSBS released the framework on September 16, 2026. It has four parts: a Core Examiner Guide with scoping questions and a document request list, an Examiner Work Program, Nonbank AI Supplements focused on third-party and vendor risk, model risk, and consumer protection, and an AI Use Case Risk Tiering Worksheet. CSBS describes it as a discretionary supervisory tool, not a new set of legal requirements, but in practice the document request list is a preview of what state examiners will ask for. Treat it as an exam-readiness checklist, and remember that state mortgage regulators (AARMR) are evaluating it for the nonbank industry.

  1. What documentation do we need ready for a GSE request or state exam?
    Keep the inventory, the tiering decisions, the approvals, the vendor assessments, the monitoring results, and the incident logs in one place — and keep them current. Both GSEs can request this information. A practical binder contains: the AI inventory with tier assignments, governance policy and committee minutes approving the policy, vendor AI questionnaires and contract language for AI/ML products, monitoring results for high-tier use cases with defined cadence and thresholds, fair lending testing for high-tier tools that affect credit decisions, incident logs for any model error or bias finding, and a current attestation or sign-off from the accountable executive.
  2. How do we handle the “less discriminatory alternative” question for AI tools?
    If a state or federal examiner challenges an AI-driven decision, you may be asked to show that there is no less discriminatory alternative that would serve your legitimate interest. Document the alternatives you considered when adopting or changing any model that affects credit decisions, pricing, marketing, or servicing, and record why you chose the one you did. This is a documentation discipline, not a separate workstream. Build it into your model change management process: when a model goes live, when it is materially updated, or when a new use case is approved, the file should include the alternatives considered and the rationale. State laws are also moving — Colorado’s Department of Law published draft automated decision-making technology rules on August 11, 2026, with a public hearing scheduled for October 26 (verify current status).

The Most Common Failure


10. What is the most common AI governance failure you see in mortgage lender programs?
Treating the governance program as a documentation exercise rather than an operational one. A written policy that no one updates, an inventory that misses half the systems, vendor questionnaires that ask the wrong questions, tiering that nobody has documented the reasoning for, monitoring that exists in a slide deck but not in the production environment. The September 2026 survey found 87% of respondents have written AI policies, but only 58% monitor tools after deployment, and only about 40% of smaller lenders do. A policy without monitoring evidence will not hold up to a GSE request or state exam. The fix is not a longer binder. It is a working program with named owners, defined cadences, and evidence of operation in the files an examiner can pull.

Want help building an AI inventory, designing vendor questionnaires, or standing up monitoring that satisfies Fannie Mae, Freddie Mac, and state examiners? Synergy works with mortgage QC and compliance teams nationwide. Book a 30-minute call to map out your program.

Disparate Impact Isn’t Dead: Why State Fair Lending Rules Now Drive Your Testing

On September 3, 2026, the Mortgage Bankers Association sued New Jersey in federal court to block the state’s disparate impact rules, and the lawsuit captures the fair lending problem lenders face going into 2027. Federal agencies have spent the past year dismantling disparate impact as an enforcement theory. Some states are moving the opposite direction and writing it into regulation. State disparate impact rules now create real liability even where federal exposure has narrowed, and any lender that scales back statistical fair lending testing on the strength of federal headlines alone is taking a risk it may not be able to see.

Here is the current landscape and how to calibrate your program.

The Federal Retreat, in Five Steps

The federal shift has been broad and fast:

  1. DOJ. On December 9, 2025, the Department of Justice published a final rule rescinding the disparate impact provisions of its Title VI regulations.
  2. HUD. On January 14, 2026, HUD proposed to remove its Fair Housing Act discriminatory effects regulation at 24 CFR 100.500 entirely, leaving disparate impact questions to the courts. Comments closed February 13, 2026 (verify whether a final rule has issued).
  3. CFPB. The CFPB’s amended Regulation B, issued April 22, 2026 and effective July 21, 2026, removed disparate impact liability from its ECOA framework, narrowed the discouragement standard, and restricted for-profit Special Purpose Credit Programs.
  4. CFPB, again. In June 2026, the CFPB rescinded its December 2020 advisory opinion on Special Purpose Credit Programs as inconsistent with the amended Regulation B.
  5. FTC. In August 2026, the FTC announced it was abandoning disparate impact and unfair discrimination theories in its consumer protection work (verify scope before relying on it).

The direction is clear. But three facts limit how far it reaches.

The Fair Housing Act has not changed

Removing a regulation does not amend a statute. In Texas Department of Housing and Community Affairs v. Inclusive Communities Project (2015), the Supreme Court held that disparate impact claims are cognizable under the Fair Housing Act, subject to safeguards. HUD’s proposal itself says it is leaving interpretation to the courts, not eliminating the theory. Private plaintiffs and fair housing organizations can still bring FHA disparate impact claims, and residential mortgage lending sits squarely within the FHA.

The amended Regulation B is being challenged

Advocacy groups and private companies filed suit in mid-2026 challenging the revised ECOA rule, including the Special Purpose Credit Program provisions. A court ruling could change the federal baseline again.

Policy can reverse

Disparate impact policy has flipped with nearly every administration since 2013. A program built only to today’s federal posture may need to be rebuilt in a few years. Testing infrastructure that is cheap to maintain is far easier to keep than to rebuild.

New Jersey: The Clearest Example of State Disparate Impact Rules

In December 2025, the New Jersey Division on Civil Rights adopted rules codifying disparate impact discrimination under the New Jersey Law Against Discrimination. The rules cover employment, housing, places of public accommodation, contracting, and housing financial assistance, which includes mortgage lending. The state’s press release was explicit that federal rollbacks do not change state law standards.

According to the MBA’s complaint, filed in the U.S. District Court for the District of New Jersey, the rules:

  • Treat a facially neutral practice as a violation if it actually or predictably results in a disproportionately negative effect on a protected class, even if the effect is not substantial or statistically significant, and even before the practice is implemented.
  • Allow a challenger to establish disparate impact using national, state, and local statistics, census data, and survey data, rather than requiring a disparity among the lender’s own applicants.
  • Place the burden on the lender, in housing and housing financial assistance, to prove there is no less discriminatory alternative that would serve its legitimate interest. In effect, the lender must prove a negative.

The MBA argues the rules lack the safeguards Inclusive Communities requires, violate the Equal Protection Clause, and are preempted by the FHA and ECOA. It seeks declaratory relief and a permanent injunction, or alternatively invalidation of the housing and housing financial assistance provisions.

The complaint also makes a point every multistate lender should note: compliance may require New Jersey-specific underwriting, pricing, and servicing policies that differ from what the lender uses elsewhere. That is the operational cost of a fragmented fair lending map.

Until a court rules, New Jersey’s rules are in effect. Lenders originating or servicing loans on New Jersey properties should treat them as binding.

What This Means for Your Fair Lending Program

The federal retreat changes your risk weighting. It does not remove the need for evidence. Here is a practical calibration.

Keep statistical testing, and segment it by state

If you lend in New Jersey or any state with an active civil rights enforcement posture, run your pricing, underwriting, and exception analyses at the state level, not only in aggregate. A disparity that disappears in a national dataset can be obvious in a single state. Maintain regression-based pricing analysis that controls for legitimate credit and loan characteristics, and keep the methodology documented so it can be reproduced.

Document business justification before you need it

Under New Jersey’s framework, a lender may have to show that a challenged practice is necessary to achieve a substantial, legitimate, nondiscriminatory interest. Write that justification down for your credit overlays, minimum loan amounts, pricing adjustments, compensation plans, and marketing footprint now. Contemporaneous documentation is far more persuasive than an explanation assembled after a complaint arrives.

Build a less discriminatory alternative review into change management

When you add or change an overlay, a pricing grid, or a model, record the alternatives you considered and why you chose the one you did. That habit directly addresses the burden New Jersey’s rules place on lenders, and it is good model governance regardless of jurisdiction.

Watch discretion closely

Pricing exceptions, fee waivers, and underwriting exceptions are where disparities most often arise because they rely on individual judgment. Require documented reasons for each exception, review them monthly, and test them by protected class where data allows. This is often the highest-value fair lending control a lender can run.

Keep intentional discrimination controls sharp

The amended Regulation B still prohibits disparate treatment. Redlining analysis, marketing reviews, comparative file reviews, and complaint monitoring remain core expectations under every federal and state framework.

Extend testing to models and vendors

If a vendor’s model or automated tool influences underwriting, pricing, or marketing decisions, its outputs belong in your fair lending analysis. Ask vendors what variables their tools use, whether they have tested for disparities, and whether they can support a less discriminatory alternative review. A disparity created by a vendor’s model is still your disparity under the Fair Housing Act and under New Jersey’s rules.

Track the litigation calendar

Two cases matter most: the MBA’s challenge to New Jersey’s rules, and the challenge to the amended Regulation B. Either could change your obligations quickly. Assign someone to monitor both and brief your compliance committee when rulings issue.

A Right-Sized Approach

Fair lending programs built during the 2021 to 2024 enforcement peak may be larger than today’s federal environment requires, and trimming low-value work is reasonable. The mistake is cutting the analytics that let you see your own disparities. Those analyses protect you in state examinations, private litigation, and the next federal policy shift. They also tell you whether your own practices are fair, which is the point.

Our fair lending compliance assessment is designed to right-size testing to your footprint, including state-level segmentation for jurisdictions like New Jersey. For the federal background, see HUD’s proposed rule on its disparate impact standard in the Federal Register.

Frequently Asked Questions

If the CFPB removed disparate impact from Regulation B, can we stop disparate impact testing?

Not safely. The Fair Housing Act still supports disparate impact claims under Inclusive Communities, private plaintiffs can still sue, and states such as New Jersey have codified disparate impact under their own laws. Reduce scope thoughtfully if warranted, but keep the statistical testing that shows you where disparities exist.

Do New Jersey’s disparate impact rules apply to out-of-state lenders?

The rules apply to housing financial assistance covered by the New Jersey Law Against Discrimination, which generally reaches lending on New Jersey properties and to New Jersey consumers. Confirm your specific exposure with counsel, but any lender with meaningful New Jersey volume should assume the rules apply while the MBA lawsuit is pending.

What is the “less discriminatory alternative” burden?

Under New Jersey’s rules, once a lender shows a challenged practice serves a substantial, legitimate, nondiscriminatory interest, the lender must also show there is no less discriminatory way to achieve that interest. Documenting the alternatives you considered when adopting a policy is the most practical way to prepare.

Has HUD finalized the removal of its disparate impact rule?

HUD proposed removing its discriminatory effects regulation on January 14, 2026, with comments due February 13, 2026. Verify whether a final rule has issued before relying on it. Either way, the Fair Housing Act itself and the Supreme Court’s Inclusive Communities decision still stand.

Want an outside view of whether your fair lending testing fits your actual risk, state by state? Synergy helps mortgage lenders and credit unions calibrate fair lending programs. Book a 30-minute call to start the conversation.

Mortgage AI Governance: Fannie, Freddie, and State Examiners Now Expect a Program

Mortgage AI governance stopped being a best practice in 2026 and became a condition of doing business. Freddie Mac’s AI requirements took effect March 3, Fannie Mae’s took effect August 6, and on September 16 the Conference of State Bank Supervisors gave state examiners a structured framework for asking how you use AI. Two weeks later, a joint survey from state mortgage regulators and the Mortgage Bankers Association showed where the industry is weakest: most lenders have an AI policy, but far fewer actually monitor their AI tools once they go live.

If your institution uses AI anywhere in origination, servicing, or quality control, including inside a vendor’s platform, you now answer to three layers of oversight. Here is what each one requires and how to close the gaps before someone asks.

Layer One: The GSE Requirements Are Already in Force

Freddie Mac Guide Section 1302.8

Freddie Mac moved first. Section 1302.8 of the Seller/Servicer Guide, added through Bulletin 2025-16 and effective March 3, 2026, requires Seller/Servicers that use AI or machine learning to maintain documented governance, establish executive accountability, implement risk management practices, and provide information about their AI safeguards when Freddie Mac requests it.

Fannie Mae Lender Letter LL-2026-04

Fannie Mae published Lender Letter LL-2026-04, Governance framework on use of artificial intelligence and machine learning, on April 8, 2026, with an effective date of August 6, 2026. If you use AI/ML in connection with loans sold to or guaranteed by Fannie Mae, or loans you service for Fannie Mae, you must:

  • Maintain written policies and procedures covering the AI/ML life cycle that reflect an understanding of applicable legal and regulatory requirements and sound risk management.
  • Comply with the Fannie Mae Information Security and Business Resiliency Supplement.
  • Manage the risks of subcontractor and vendor AI/ML use with governance that is no less protective than your own requirements.
  • Promptly disclose, on request, the types of AI/ML you use, the purpose and manner of use, the safeguards you have implemented, and anything else Fannie Mae requires.

Two points deserve emphasis. First, neither GSE requires you to adopt AI. The requirements apply only if you use it. Second, “use” is broader than most lenders assume. An automated document classification engine, an income calculation tool, a chatbot answering borrower questions, a servicing call-summary tool, or a QC platform that flags defects with machine learning all count. If you are approved by both GSEs, you must satisfy both sets of requirements, and they are not identical (verify current text of each).

The enforcement mechanism is familiar: a loan that does not meet Guide requirements is a repurchase risk.

Layer Two: State Examiners Now Have an AI Playbook

On September 16, 2026, CSBS released its Artificial Intelligence Supervisory Framework for state financial examiners. It has four parts:

  • Core Examiner Guide: scoping questions, a document request list, and procedures covering AI governance and oversight, AI inventories and use cases, and generative AI and other emerging uses.
  • Examiner Work Program: detailed guidance for applying the core guide.
  • Nonbank AI Supplements: focused on third-party and vendor risk, model risk, and consumer protection.
  • AI Use Case Risk Tiering Worksheet: an optional tool for assessing individual AI applications and deciding whether deeper review is warranted.

CSBS describes the framework as a discretionary supervisory tool, not a new set of legal requirements, and each state agency will decide how much of it to adopt. Do not read that as permission to wait. State mortgage regulators, through the American Association of Residential Mortgage Regulators (AARMR), are evaluating it for the nonbank mortgage industry, and multistate exam tools tend to spread quickly once adopted. The document request list is effectively a preview of what you will be asked to produce.

There is also a federal gap the framework fills. The banking agencies’ 2026 revisions to model risk management guidance expressly exclude generative and agentic AI, so for many institutions the CSBS framework is the most specific supervisory roadmap available.

Layer Three: The Industry’s Own Report Card

On September 28, 2026, AARMR, the MBA, and Boston Consulting Group released The State of AI Among Mortgage Lenders and Servicers. The respondents were roughly 30 lenders and servicers representing about 40% of the market. Key findings:

  • 87% have written AI policies or standards, but only 58% reported ongoing monitoring for issues such as accuracy and model drift.
  • Among smaller lenders, only about 40% reported ongoing monitoring, compared with about 80% of larger lenders.
  • 59% named regulatory and compliance uncertainty as their biggest barrier to scaling AI, ahead of unclear return on investment (45%) and data quality (24%).
  • Only about a quarter had scaled at least one AI use case, and adoption is concentrated in productivity and document-processing work.

Read those numbers as an examiner would. The weakest point in mortgage AI governance is not the policy. It is what happens after deployment. Examiners and GSE reviewers will ask to see evidence of monitoring, not just a policy that promises it.

Building a Mortgage AI Governance Program That Holds Up

You do not need a large data science team to meet these expectations. You need a disciplined program with six components.

1. An AI inventory you can produce on demand

List every AI or machine-learning use case, internal or vendor-supplied. For each one, capture the business function, the owner, the vendor (if any), what data it consumes, whether it touches nonpublic personal information, what decision or output it influences, and whether a human reviews the output before it affects a borrower. Both GSEs and the CSBS framework start here. If you cannot produce the list in a day, that is your first finding.

2. Risk tiering

Not every use case deserves the same scrutiny. A tool that summarizes internal policy documents is low risk. A tool that calculates qualifying income, influences pricing, decides which files get escalated, or drafts adverse action language is high risk because it can affect credit decisions, fair lending outcomes, or required disclosures. The CSBS risk tiering worksheet is a reasonable starting template. Document the tier and the reasoning.

3. Accountability at the executive level

Name an accountable executive and a governance forum that approves new use cases, reviews tier assignments, and receives monitoring results. Record it in board or committee minutes. “IT handles it” will not satisfy Freddie Mac’s executive accountability expectation.

4. Vendor oversight that is “no less protective”

This is where most lenders are exposed, because most mortgage AI arrives inside vendor products. Update vendor due diligence questionnaires to ask whether the product uses AI/ML, for what, how it is tested and monitored, what data trains it, where that data is processed, and how errors are reported. Add contract language requiring notice before material model changes and cooperation with GSE or regulator information requests. Fannie Mae’s “no less protective” standard means a vendor’s AI is your AI for governance purposes.

5. Post-deployment monitoring

This is the gap the survey exposed. For each high-tier use case, define what accuracy means, how often you test it, and what threshold triggers escalation. For document extraction and income calculation tools, the cleanest control is QC sampling: compare AI outputs to human-verified results on a defined sample each month and trend the error rate. Your existing QC function is already built for this kind of testing. Use it.

6. Records

Keep the inventory, tiering decisions, approvals, vendor assessments, monitoring results, and incident logs in one place. Both GSEs can request this information, and CSBS’s document request list assumes it exists.

The Fair Lending and Disclosure Overlay

AI governance is not a separate silo. Any tool that influences underwriting, pricing, or adverse action still has to work within ECOA and Regulation B, including the requirement to give specific principal reasons for adverse action. If a model contributes to a denial, you must still be able to explain the reasons in terms the borrower can understand. State laws are moving too. Colorado’s Department of Law published draft automated decision-making technology rules on August 11, 2026, with a public hearing scheduled for October 26 (verify current status). Build your program so it can absorb state requirements without a rebuild.

Where to Start This Quarter

If you are behind, sequence the work: inventory first, tiering second, vendor questionnaires third, monitoring for your top three high-tier use cases fourth. That order gives you something defensible to show within 60 days. Our compliance services team can help you build the inventory and design QC-based monitoring that fits your existing review cycle.

Frequently Asked Questions

Do the GSE AI requirements apply if we only use AI through vendors?

Yes. Fannie Mae requires you to manage vendor and subcontractor AI/ML risk with governance no less protective than its requirements, and Freddie Mac’s Section 1302.8 applies to your use of AI or machine learning regardless of who built it. If a vendor’s tool touches loans you sell to or service for the GSEs, it belongs in your inventory and your oversight program.

Is the CSBS AI Supervisory Framework binding?

No. CSBS calls it a discretionary supervisory tool, and each state decides how to incorporate it. In practice, it tells you what state examiners are likely to ask for, so treat its document request list as an exam-readiness checklist.

What is the most common gap in mortgage AI governance?

Ongoing monitoring. The September 2026 AARMR, MBA, and BCG survey found 87% of respondents had written AI policies but only 58% monitored tools after deployment, with smaller lenders at roughly 40%. A policy without monitoring evidence will not hold up in a GSE review or state exam.

Does using AI change our adverse action obligations?

No. ECOA and Regulation B still require specific principal reasons for adverse action. If an AI tool contributes to a credit decision, you need to be able to identify and explain those reasons accurately, which is one reason high-impact tools belong in your highest risk tier.

Need help building an AI inventory or designing monitoring that satisfies Fannie Mae, Freddie Mac, and state examiners? Synergy works with mortgage QC and compliance teams nationwide. Book a 30-minute call to map out your program.

Web Statistics