Archives

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.

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.

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.

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.

Mortgage AI/ML Compliance Q&A: Governance, Fair Lending, and Exam Readiness

AI/ML compliance for mortgage lenders is no longer a future-state planning exercise. Fannie Mae’s LL-2026-04 took effect August 6, 2026. Freddie Mac’s Section 1302.8 took effect March 3, 2026. The Colorado AI Act took effect February 1, 2026. The California DFPI is actively examining AI/ML programs. State AGs are bringing actions under existing UDAP authority.

This Q&A focuses on the practical questions your compliance team will face as the AI/ML governance framework comes into operational reality in 2026 — what to build, what to document, what examiners are looking at, and how to keep the program running as models evolve and rules change.

Q1: We Just Discovered LL-2026-04. What Do We Do First?

The first step is the AI/ML use case inventory. You cannot scope a governance program without knowing which systems are in scope. The inventory should identify every AI/ML system used in the mortgage lifecycle, the business function each supports, the data inputs, the model owner, the deployment date, and the underlying vendor (if third-party).

A common mistake is to start with the governance policy. A policy written before the inventory is complete is a policy that does not match the actual system footprint. The inventory drives the policy, not the other way around.

Aim to have a complete inventory within 30 days. Most lenders underestimate how many AI/ML systems they actually run — the inventory typically surfaces 30-50% more systems than the compliance team expected.

Q2: How Do We Decide Which AI/ML Systems Are “High-Impact”?

The GSE frameworks and the Colorado AI Act both use a risk-tiering approach. The relevant question for a high-impact designation is whether the system can materially affect a borrower’s loan terms, access to credit, or experience in the loan process.

High-impact systems for mortgage lenders typically include:

  • Automated underwriting systems (AUS)
  • AI-driven appraisal valuation models (AVMs with machine learning components)
  • AI-driven fraud detection models that affect application decisions
  • Pricing optimization models that influence loan pricing
  • Lead scoring models that affect credit decisions
  • Income and asset verification tools that use AI
  • Customer service chatbots that handle credit-related inquiries

Lower-risk systems include internal marketing analytics, business intelligence dashboards, and back-office automation that does not affect borrower outcomes. Document the risk tiering methodology so examiners can review the logic.

Q3: What Counts as an “AI/ML System” for Compliance Purposes?

The LL-2026-04 definition tracks the broad industry usage. An AI/ML system is any system that uses statistical learning, neural networks, natural language processing, or other machine learning techniques to produce an output from training data. Generative AI (text or image generation), predictive models, classification models, and clustering models are all in scope.

Out of scope: rule-based decision engines that do not learn from data, simple threshold-based scoring (e.g., credit score lookups without model adjustment), and traditional statistical models without a learning component. The key question is whether the system’s parameters are learned from data rather than set by human designers.

When in doubt, include the system in the inventory and document the determination. The cost of including a borderline system is much lower than the cost of an examiner discovering a missed system.

Q4: How Do We Build a Fair Lending Testing Program for AI/ML?

A defensible fair lending testing program has four components.

1. Outcomes-based testing. Compare actual loan decisions, pricing, or other outcomes across demographic segments. The relevant segments under federal law are race, national origin, sex, religion, familial status, age, and disability. Under state law, additional protected categories may apply.

2. Input-based testing. Audit model features for proxy variables that correlate with protected classes even when the protected class is not a direct input. ZIP code is a classic proxy for race. Language preference can be a proxy for national origin. Proxies are not always a violation, but they require documentation of why the proxy is appropriate and how its use is monitored.

3. Segment-level performance review. Model accuracy, false positive rates, and false negative rates should be reviewed at the segment level. A model that performs well on average but has materially different error rates across protected segments is a model that needs remediation before deployment.

4. Counterfactual testing. For loan decisions, change the protected class of an applicant and observe whether the decision changes. If the decision changes when only the protected class changes, the model is using protected class or proxy information inappropriately.

The testing should be performed before deployment, after material model changes, and at a defined cadence — at least annually for high-impact systems.

Q5: What Documentation Do Examiners Look For First?

Examiners start with the inventory and the governance policy. From there, the document request typically expands to model documentation, validation reports, fair lending testing, vendor contracts, and the most recent attestation.

The most common finding in early AI/ML examinations is a gap between what the inventory claims and what the documentation actually supports. Lenders may claim to have a fair lending testing program but produce a single slide with results, not a documented testing protocol with methodology, results, and remediation actions.

The exam-ready binder for AI/ML governance should include:

  • AI/ML use case inventory with risk tiering
  • Governance policy approved at senior committee level
  • Model documentation for each high-impact system (data sources, training methodology, performance metrics, validation results)
  • Fair lending testing reports for each high-impact system
  • Vendor contracts with AI/ML-related terms
  • Annual attestation
  • Incident log (any model failures, complaints, regulatory inquiries)

Q6: How Do We Handle AI Tools That Loan Officers Use Independently?

If loan officers use AI tools (ChatGPT, Claude, specialized mortgage AI assistants, etc.) in connection with loan files, the tools are in scope under LL-2026-04 and Section 1302.8. The lender is the deployer and is accountable for the tool’s compliance.

The minimum controls: an approved list of AI tools that may be used in connection with loans, prohibition on uploading borrower non-public personal information to tools that have not been approved, a confidentiality review of the tool’s data handling practices, and a documented training program for loan officers on the approved-use policy.

The most common exam finding: lenders have no visibility into which AI tools loan officers are using. Shadow AI use is a significant risk, and lenders are expected to address it proactively.

Q7: How Should We Structure the AI Governance Committee?

The committee structure varies by institution size. For mid-size and large lenders, the typical structure is:

AI Governance Committee: senior leadership (CRO, CIO, General Counsel, Head of Compliance, Head of Model Risk) meets quarterly or more frequently. Approves the governance policy, reviews high-impact model changes, signs off on attestations.

Model Risk Management function: dedicated staff (or a vendor) that runs the inventory, conducts validation, performs fair lending testing, maintains documentation.

Model Owners: business line leaders responsible for individual AI/ML systems. Own the system lifecycle, escalate issues to the committee, ensure documentation is current.

For smaller lenders without dedicated model risk staff, the model risk function may be outsourced or combined with compliance. The committee structure remains the same; the execution is shared.

Q8: How Do We Balance AI Innovation with AI Compliance?

The right framing is not “innovation vs. compliance” — it is “innovation with governance.” A model that cannot be explained to an examiner is a model that creates regulatory risk. A model that produces disparate outcomes is a model that creates litigation risk. Governance is what makes innovation sustainable.

The practical implementation: the governance framework should be designed to support business velocity, not slow it down. Pre-deployment validation, fair lending testing, and documentation should be efficient and well-scoped. The model risk function should be a partner to the business, not a bottleneck.

A common failure mode: over-engineering the governance process to the point where business teams route around it. The result is shadow AI use — business teams adopt new tools without governance review, and the compliance program loses visibility into the actual system footprint.

Q9: What Is the Most Common AI/ML Compliance Failure You See?

The most common failure is treating the AI/ML governance program as a documentation exercise rather than an operational one. Lenders produce a policy and a checklist, but they do not actually run the inventory, conduct the validation, or perform the fair lending testing. When the examiner asks for the supporting documentation, the program collapses.

The second most common failure is treating vendor-provided validation as a substitute for lender validation. The lender is accountable for the model’s performance in its own use context. The vendor’s validation report is an input, not an output.

The third most common failure is fair lending testing that is too narrow — testing only adverse action outcomes and missing proxy variable analysis, or testing only protected classes under federal law and missing the additional state-level protected categories.

Q10: Where Should AI/ML Compliance Be on the Q3 2026 Priority List?

For lenders that have not yet built a program, AI/ML governance should be at the top of the Q3 2026 priority list. The LL-2026-04 effective date has passed, and the first attestation cycle is approaching. The work cannot wait for Q4.

The Q3 priorities, in order:

This month: Complete the AI/ML use case inventory. Identify model owners and risk tiering.

Next 30 days: Draft the governance policy. Get committee approval.

Next 60 days: Begin fair lending testing for the highest-impact systems. Document the testing methodology.

By year-end: Complete the first round of validation. Stand up the annual attestation process. Build the exam-ready binder.

For lenders with a program already in place, the Q3 priority is to harden the documentation for examiner review and to verify the program covers the new state-level rules — particularly the Colorado AI Act if you originate or service in Colorado.

Need support on AI/ML governance, fair lending testing, or state-level compliance overlay? Synergy works with mortgage lenders on AI/ML program design, validation, multi-state compliance overlays, and exam readiness. Book a 30-minute AI/ML review.

State AI/ML Enforcement for Mortgage Lenders: What You Need to Know in 2026

Federal AI/ML guidance for mortgage lending remains a patchwork. The CFPB has issued interpretive guidance, the GSEs have published AI governance frameworks, and Congress has considered — but not passed — federal AI legislation. In the absence of a unified federal standard, state regulators have moved ahead with their own rules.

For mortgage lenders operating in multiple states, the practical effect is a growing set of state-specific AI/ML obligations layered on top of federal and GSE requirements. This article walks through the state rules that affect mortgage lenders in 2026, how they interact with each other and with federal frameworks, and what a defensible multi-state AI compliance program looks like.

Why State AI/ML Regulation Matters for Mortgage Lenders

State AI/ML rules were not written with mortgage lending as the primary use case. Most originate in consumer protection, employment, or insurance contexts. But the definitions of “automated decision tool,” “high-risk AI system,” and “consumer” are broad enough to capture mortgage lending activity — and state regulators have signaled that they will apply their AI rules to financial services, not just to the originally-targeted industries.

The compliance exposure is not theoretical. State AGs have been active in 2025 and 2026, with several settlements against financial services firms involving AI/ML use. Mortgage lenders that treat state AI compliance as a low-priority “tech company” issue are misreading the landscape.

Colorado AI Act (SB 24-205)

The Colorado AI Act took effect on February 1, 2026. It is the most comprehensive state AI statute in the United States and the one most likely to set a template for other states.

Who It Applies To

The Act applies to “developers” and “deployers” of “high-risk AI systems” used in Colorado. A deployer is any entity that uses a high-risk AI system to make decisions that affect Colorado residents. Mortgage lenders that use AI/ML systems for Colorado residents — including in origination, servicing, marketing, or customer service — are deployers under the Act.

What Counts as a High-Risk AI System

The Act specifies categories of high-risk systems. The relevant categories for mortgage lenders include:

  • AI systems used to make decisions about access to financial services, including credit
  • AI systems used for employment decisions (relevant for HR, recruiting, and internal loan officer evaluation)
  • AI systems that materially affect access to housing

Automated underwriting systems, AI-driven pricing tools, lead scoring systems that influence credit decisions, and AI-driven appraisal valuation models are all high-risk under the Act.

What Deployers Must Do

Deployers must implement a risk management program and policy, complete an impact assessment for each high-risk system, provide notice to consumers that an AI system is being used, and allow consumers to request human review of an AI-driven decision. The impact assessment must be made available to the Colorado AG on request.

The Act also prohibits algorithmic discrimination — defined in ways that overlap significantly with federal fair lending law but include additional categories of protected activity.

California AI Rules

California does not yet have a comprehensive AI statute, but the state has a layered set of AI-related rules that affect mortgage lenders.

AB 2013 (Generative AI Training Data)

AB 2013 requires developers of generative AI systems to publish a summary of the training data used. The rule is targeted at generative AI providers, not at deployers. For mortgage lenders, the relevance is downstream: if you use a generative AI tool (e.g., for document drafting, customer service), the underlying provider’s compliance affects your vendor risk profile.

SB 942 (AI Transparency)

SB 942 requires AI providers to offer a free AI detection tool to users. The rule is targeted at AI providers, not at deployers. The relevance is the same as AB 2013 — vendor due diligence.

California Department of Financial Protection and Innovation (DFPI)

The California DFPI has been the most active state financial regulator on AI/ML. The DFPI has issued multiple guidance documents on AI use in lending, conducted examinations focused on AI/ML systems, and entered into consent orders with lenders involving AI/ML model risk management and fair lending testing.

The DFPI’s approach is consistent with the federal and GSE frameworks, but it includes California-specific requirements on consumer notification, model documentation, and the right to human review. For lenders operating in California, DFPI expectations are effectively a fourth layer of AI/ML governance on top of CFPB, GSE, and other state rules.

New York State and City

New York has not passed a comprehensive state AI statute, but the New York Department of Financial Services (DFS) has issued guidance on AI/ML use by regulated financial institutions. The DFS guidance is supervisory in nature — not a binding regulation — but it sets the expectation for AI/ML governance programs for insurers and banks under DFS jurisdiction.

For mortgage lenders operating in New York, the DFS guidance effectively requires an AI/ML governance program consistent with the GSE frameworks. Examiners will look for documentation of model risk management, fair lending testing, and consumer protection controls.

New York City Local Law 144 (automated employment decision tools) affects mortgage lenders that use AI in hiring — for loan officer recruiting, underwriting staff screening, or any other employment decision. The law requires an annual bias audit and public posting of the audit results.

Texas: UDAP Authority

Texas has not passed a state AI statute, but the Texas Attorney General and state financial regulators have used existing Unfair, Deceptive, or Abusive Acts or Practices (UDAP) authority to bring AI/ML-related actions. The state has been particularly active on AI-driven decisions that result in disparate impact on protected classes — using UDAP rather than a separate AI statute.

For Texas-licensed mortgage lenders, the practical implication is that AI/ML use is regulated — even without a specific AI statute. The compliance program that satisfies the Texas SML for the SSSF, the GSE AI/ML frameworks, and the federal fair lending rules is the foundation for UDAP defensibility.

Other State Activity

State AI/ML activity is moving fast. The states that have either passed AI rules or have active rulemaking in 2026 include Illinois, New Jersey, Virginia, Washington, and Oregon. The rules vary in scope and approach, but the direction is consistent: more state regulation, with mortgage lending in scope.

A practical approach for mortgage lenders operating nationally: design the AI/ML governance program to the strictest applicable state requirement, and apply it uniformly. The marginal cost of running a single, conservative program is much lower than the cost of running multiple state-specific programs.

How State AI Rules Interact with Federal and GSE Frameworks

A consolidated view of the AI/ML compliance landscape for mortgage lenders in 2026:

  • CFPB: interpretive guidance, focus on adverse action notices, fair lending, and accuracy of AI-driven decisions. Exam focus in 2026.
  • Fannie Mae LL-2026-04: AI/ML governance framework, effective August 6, 2026. Six governance obligations, annual attestation.
  • Freddie Mac Section 1302.8: companion AI/ML governance framework, effective March 3, 2026. Substantively similar to LL-2026-04.
  • Colorado AI Act: high-risk AI system requirements, impact assessments, consumer notice, right to human review. Effective February 1, 2026.
  • California DFPI: AI/ML supervisory guidance, focused on documentation, fair lending testing, and consumer protection.
  • New York DFS: supervisory guidance, treated as effective standard for New York-licensed institutions.
  • State AGs: UDAP authority, used to bring AI/ML-related actions in states without specific AI statutes.

These frameworks are additive. A lender that satisfies Fannie Mae LL-2026-04 still has separate Colorado, California, and New York obligations. The Colorado impact assessment is not a substitute for the LL-2026-04 attestation.

Building a Multi-State AI/ML Compliance Program

A defensible program has five components.

1. Unified AI/ML Use Case Inventory

Maintain a single inventory that flags which systems are in scope under which state rules. The inventory is the foundation — without it, you cannot determine your compliance obligations.

2. State-Overlay Documentation

For each state with AI/ML rules, maintain an overlay document that maps the state requirements to your existing governance program. The overlay identifies the gaps and the remediation work.

3. Consumer Notice and Human Review Workflows

Colorado and other state rules require consumer notice of AI use and a right to human review. The workflows should be designed to comply with the strictest applicable state requirement.

4. Impact Assessment Library

Maintain an impact assessment for each high-risk AI system, updated annually or after material model changes. The assessment is a state regulatory document, not a federal one — different states may require different formats.

5. Multi-State Vendor Oversight

Your vendor oversight program must cover state-specific requirements. A vendor providing AI/ML services in Colorado has different disclosure obligations than a vendor providing the same services in Texas.

Frequently Asked Questions

Does the Colorado AI Act Apply If We Don’t Have a Physical Office in Colorado?

Yes. The Act applies to any deployer that uses a high-risk AI system to make decisions affecting Colorado residents. If you originate or service a mortgage for a Colorado resident and use an AI/ML system in connection with that loan, the Act applies.

How Do State AI Rules Interact with Fair Lending Law?

State AI rules typically add anti-discrimination requirements that overlap with federal fair lending law but are not identical. A fair lending test that satisfies the CFPB may not satisfy the Colorado AI Act. The conservative approach is to test to the strictest applicable standard.

What If a Vendor Provides Our AI/ML System? Are We Still Liable?

Yes. Under the Colorado AI Act, the GSE frameworks, and most state-level approaches, the lender is the deployer and remains accountable for the system’s compliance. Vendor contracts can shift some financial risk, but not regulatory risk.

What Records Must We Retain?

Three years for most state requirements, but some state rules require longer retention for impact assessments. Document the retention requirement for each state in scope and apply the longest applicable period uniformly.

Need help designing or auditing your multi-state AI/ML compliance program? Synergy supports mortgage lenders with state overlay documentation, impact assessment design, and multi-state governance program reviews. Book a 30-minute program review.

Fannie Mae AI/ML Governance: What Lenders Must Do by August 6

Fannie Mae’s AI/ML governance framework — Lender Letter LL-2026-04 — took effect on August 6, 2026. For any single-family seller or servicer using artificial intelligence or machine learning in connection with mortgages sold to Fannie Mae, the framework is now in force. The compliance bar is no longer aspirational; it is operational.

LL-2026-04 is the companion to Freddie Mac’s Seller/Servicer Guide Section 1302.8, which took effect March 3, 2026. Together, the two frameworks establish the GSE position on AI/ML governance: lenders are accountable for the design, performance, and outcomes of any AI/ML system used in the mortgage lifecycle, regardless of whether the system is built in-house, provided by a third-party vendor, or accessed through a marketplace platform.

This guide walks through what LL-2026-04 requires, how it interacts with Freddie Mac Section 1302.8 and state-level AI rules, and what a defensible AI/ML governance program looks like for a mortgage lender as of August 2026.

What LL-2026-04 Actually Requires

LL-2026-04 is structured around six governance obligations. Each is a stand-alone compliance topic and each requires documentary evidence.

1. AI/ML Use Case Inventory

Lenders must maintain a complete inventory of every AI/ML system used in the mortgage lifecycle. The inventory should identify the system, the business function it supports, the data inputs, the model owner, the deployment date, and the underlying vendor (if third-party).

In scope: automated underwriting, appraisal valuation models, fraud detection, lead scoring, marketing optimization, customer service chatbots, document classification, income and asset verification, and any pricing or margin optimization tool that uses statistical learning.

Out of scope: rule-based decision engines that do not learn from data, simple statistical scoring (e.g., credit score lookups without model adjustment), and standard business intelligence dashboards.

2. Model Risk Management Framework

Each inventoried AI/ML system must be classified by risk tier based on its impact on loan decisions, borrower outcomes, and regulatory exposure. High-impact systems (underwriting, pricing, fraud, valuations) require the most rigorous controls.

The framework should document validation activities (pre-deployment testing, ongoing monitoring, periodic revalidation), performance thresholds, change management procedures, and override mechanisms for human review.

3. Fair Lending Testing

Each AI/ML system that affects loan decisions, pricing, or adverse action notices must be tested for fair lending impact. Testing should include disparate impact analysis across prohibited basis categories (race, national origin, sex, religion, familial status, age, disability), proxy variable analysis (identifying features that correlate with protected classes even when the protected class is not a direct input), and segment-level performance review.

The testing should occur before deployment, after material model changes, and at a defined cadence (typically annually for high-impact systems).

4. Governance Documentation

LL-2026-04 requires a written AI/ML governance policy that is approved at the board or senior committee level. The policy should cover roles and responsibilities, model lifecycle controls, escalation paths, exception handling, and incident response.

Documentation must be maintained for the life of each model plus a defined retention period. The retention floor is generally three years post-decommissioning, consistent with other mortgage compliance records.

5. Third-Party Vendor Oversight

Lenders remain accountable for AI/ML systems provided by third parties. The oversight program should include vendor due diligence (model documentation review, validation access, audit rights), ongoing monitoring (performance reports, incident notification, regulatory change tracking), and contractual protections (indemnification, data security, model decommissioning rights).

A common gap: lenders that treat vendor systems as “off the shelf” and skip validation. LL-2026-04 treats this as a compliance failure. The lender is responsible for validating the model in the context of its own use, even if the vendor provides the validation methodology.

6. Annual Attestation

Lenders must attest annually to Fannie Mae that they have an AI/ML governance program in place that meets LL-2026-04 requirements. The attestation is a senior officer certification, not a procedural check-the-box. Officers signing the attestation should expect to defend the substance of the program if challenged.

How LL-2026-04 Interacts with Freddie Mac Section 1302.8

If your institution sells to both GSEs, you do not need to maintain two separate AI/ML governance programs. A unified program that satisfies both frameworks is acceptable, and Fannie Mae and Freddie Mac have signaled that they will accept each other’s attestations in most cases.

The two frameworks differ in three operational details:

  • Effective dates: Freddie Mac Section 1302.8 took effect March 3, 2026. Fannie Mae LL-2026-04 took effect August 6, 2026. If you implemented a Section 1302.8 program in the spring, you should be in good shape on the substance of LL-2026-04. The remaining work is typically attestation timing and documentation reconciliation.
  • Attestation cadence: Freddie Mac requires annual attestation. Fannie Mae requires annual attestation. The two attestations can be filed separately even if the underlying program is the same.
  • High-impact system definition: The two frameworks use slightly different definitions of “high-impact.” Where they differ, the more conservative definition should govern.

If your institution sells to only one of the two GSEs, you only need to meet that GSE’s framework. But state-level AI rules may still apply regardless of GSE relationship — see the August 2026 article on state AI/ML enforcement.

What “In Connection With Mortgages Sold to Fannie Mae” Means

LL-2026-04 applies to AI/ML systems used in connection with mortgages sold to Fannie Mae. The phrase is interpreted broadly. If a system touches a loan that may eventually be sold to Fannie Mae, the governance obligations apply.

In practice, this covers:

  • Systems used at the point of application (lead scoring, prequalification)
  • Systems used during origination (automated underwriting, fraud detection, document processing)
  • Systems used post-closing (servicing decisioning, loss mitigation, default management)
  • Marketing and customer service systems if they influence the loan pipeline that includes Fannie Mae-sold loans

A practical approach: inventory every AI/ML system in your mortgage technology stack. If any of them touch a loan that may be sold to Fannie Mae, the system is in scope.

The Fair Lending Layer

The fair lending testing requirement under LL-2026-04 is the area where most lenders are least prepared. Standard model risk management covers performance, drift, and stability. Fair lending testing is a separate discipline with its own methodology, its own tooling, and its own documentation requirements.

If your institution has not yet built a fair lending testing program for AI/ML, the August 6 effective date is the trigger to either build it or engage external support. The testing cadence is at least annual for high-impact systems, and the documentation must be available for Fannie Mae review on request.

Two common testing approaches:

  • Outcomes-based testing: Compare actual loan decisions, pricing, or other outcomes across demographic segments. Identifies disparate impact at the output level.
  • Input-based testing: Audit model features for proxies that correlate with protected classes. Identifies structural risk before the model produces an outcome.

The most defensible approach is both. Outcomes-based testing identifies what the model is doing. Input-based testing identifies how the model could produce a problematic outcome before it happens.

Action Steps for August 2026

If your institution has not yet built a LL-2026-04 program, the immediate priorities are:

This month: Inventory every AI/ML system in the mortgage technology stack. Identify model owners, deployment dates, and vendors. This is the foundation for everything else.

This quarter: Classify each system by risk tier. Document the validation activities already performed. Identify gaps relative to LL-2026-04 requirements.

By year-end: Complete the governance policy. Establish fair lending testing for high-impact systems. Build the vendor oversight program. Stand up the annual attestation process.

By Q1 2027: Complete the first attestation cycle. Document the validation work performed. Build the exam-ready binder.

Frequently Asked Questions

We Don’t Use AI/ML Anywhere. Does LL-2026-04 Still Apply?

If you genuinely use no AI/ML systems in connection with mortgages sold to Fannie Mae, the framework does not impose substantive obligations. The annual attestation, however, is still required — you attest that you have no in-scope systems. Document the basis for that conclusion (the inventory and the analysis) so the attestation is defensible.

What About AI Tools Used by Individual Loan Officers?

If a loan officer uses a third-party AI tool (e.g., a ChatGPT-style assistant) in connection with a loan, the tool is in scope. The lender’s vendor oversight program must cover the tool, including the data security and confidentiality controls.

How Does LL-2026-04 Interact With State AI Laws?

LL-2026-04 is a GSE framework. State AI laws are separate obligations. In most cases, the state law is additive — you must satisfy both. See the August 2026 article on state AI/ML enforcement for the specific state rules that apply to mortgage lenders.

What Records Must We Retain?

Three years post-decommissioning for each model, consistent with other mortgage compliance records. The retention applies to model documentation, validation results, fair lending testing, governance decisions, vendor contracts, and attestation records.

Ready to build or audit your AI/ML governance program? Synergy supports mortgage lenders with model inventory design, governance policy drafting, fair lending testing frameworks, and AI/ML exam-readiness reviews. Book a 30-minute governance review.

Fannie Mae & Freddie Mac AI Governance: What Lenders Must Do

Two GSEs, Two Deadlines, One Compliance Reality

The mortgage industry has entered a new era of AI governance — and this shift is coming from the two government-sponsored enterprises, not just federal agencies.

On March 3, 2026, Freddie Mac updated its Seller/Servicer Guide Section 1302.8, establishing formal governance requirements for any seller/servicer using artificial intelligence or machine learning in connection with mortgages sold to Freddie Mac. Then, on April 8, 2026, Fannie Mae issued Lender Letter LL-2026-04 — its own AI/ML governance framework for single-family sellers and servicers — effective August 6, 2026.

Together, these mandates create a dual-layer compliance obligation affecting virtually every lender operating in the conventional space. Whether you sell to Freddie Mac, Fannie Mae, or both, your AI governance infrastructure is now under regulatory scrutiny that didn’t exist twelve months ago.

What Freddie Mac Requires (Section 1302.8, Effective March 3, 2026)

Freddie Mac’s guidance, issued under Bulletin 2025-16, sets the baseline. Any seller/servicer using AI or ML in loan origination or servicing must establish a clear, documented governance framework covering:

  1. Policies, processes, and procedures governing AI/ML adoption and use throughout the loan lifecycle
  2. Risk management actions for AI/ML systems — how the lender identifies, assesses, and mitigates AI-related risk
  3. Senior management approval of AI/ML policies — a higher bar than simply documenting them
  4. Indemnification obligations — lenders assume full responsibility for AI-driven decisions, including outcomes from vendor-provided AI tools
  5. Compliance with applicable law and Freddie Mac Purchase Documents

The indemnification obligation deserves special attention. If a lender’s vendor supplies an AI-driven underwriting tool that produces a discriminatory outcome, the lender — not the vendor — bears accountability to Freddie Mac. Vendor due diligence is no longer optional.

What Fannie Mae Requires (LL-2026-04, Effective August 6, 2026)

Fannie Mae’s Lender Letter LL-2026-04 takes effect August 6, 2026 — giving lenders a longer runway than Freddie Mac’s mandate, but no less urgency.

The framework requires single-family sellers and servicers to maintain a documented, actively maintained AI/ML governance program including:

  1. Written policies and procedures covering the full life cycle of any AI/ML system — from development to ongoing maintenance
  2. Annual policy reviews with a designated owner responsible for implementing, maintaining, and updating policies
  3. Trustworthy and ethical AI principles incorporated into the lender’s governance approach
  4. Information security compliance per Fannie Mae’s Information Security and Business Resiliency Supplement
  5. Vendor and subcontractor governance — AI/ML risk management standards must extend to any third-party tool, with protections no less robust than the lender’s own
  6. On-demand disclosure — upon request by Fannie Mae, lenders must promptly disclose AI/ML technologies deployed, their intended purposes, and risk safeguards in place

Fannie Mae’s requirements are less prescriptive than Freddie Mac’s in some areas — but they compensate with annual review obligations, designated ownership requirements, and explicit disclosure authority that gives Fannie Mae visibility into the AI tools lenders are using.

The Overlap With the New Fair Lending Framework

The timing is not coincidental. The CFPB’s April 2026 Regulation B final rule — shifting fair lending enforcement from a disparate impact to an intent-based standard — takes effect July 21, just weeks before Fannie Mae’s August 6 deadline.

If a lender uses AI-driven underwriting or pricing models and those tools produce discriminatory outputs, intentional use of that tool could constitute intentional discrimination under the new ECOA framework. Fannie Mae’s governance framework, with its emphasis on trustworthy and ethical AI, becomes a lender’s first line of defense — evidence that AI tools were deployed responsibly.

This means your AI governance documentation is no longer just a GSE compliance obligation. It is a fair lending defense. Quality control (QC) programs need to account for this intersection — reviewing whether AI tool outputs are defensible under the new intent-based standard.

Building a Compliant AI Governance Framework: Where to Start

Inventory your AI/ML tools. Identify every AI or machine learning system used in loan origination, underwriting, pricing, or servicing — including vendor-supplied tools.

Establish written policies and procedures. Both GSE frameworks require documented policies covering the full AI/ML lifecycle, reflecting legal requirements, ethical AI principles, and your institution’s risk tolerance.

Designate an owner. Fannie Mae requires a designated owner for annual reviews; Freddie Mac requires senior management approval. These establish internal accountability — and external defensibility.

Extend governance to vendors. Lenders bear responsibility for AI tool outcomes regardless of vendor involvement. Third-party AI vendors must be subject to the same governance standards.

Integrate AI governance into your quality control program. Quality control reviews should assess AI tool outputs, document governance compliance, and verify that AI-driven decisions are defensible under the new fair lending framework.

The Bottom Line: AI Governance Is Now Mortgage Compliance

Fannie Mae and Freddie Mac have made their position clear: AI governance is not a future consideration — it is a present obligation. With Freddie Mac’s deadline already passed and Fannie Mae’s approaching, lenders without established frameworks are exposed on multiple fronts.

The stakes go beyond GSE compliance. Ungoverned AI tools can expose lenders to fair lending liability, investor repurchase risk, and regulatory enforcement — particularly as the CFPB’s intent-based fair lending standard takes effect.

At Synergy, we help mortgage lenders build and maintain AI governance frameworks that satisfy GSE requirements and hold up under regulatory scrutiny. Our QC platform integrates AI governance assessment into your broader compliance program — so you’re covered on every front.

Need help building or reviewing your AI governance framework before the August 6 deadline? Contact Synergy to speak with a compliance specialist, or book a demo at simplifyqc.com.

Web Statistics