Fannie Mae AI Governance Requirements Are Now in Effect
Move From AI Policy to Provable Execution Control
Fannie Mae's AI and machine-learning governance requirements make policies, risk management, vendor oversight, and safeguard disclosure immediate priorities for covered Single-Family sellers and servicers.
Can you prove your AI had authority before it acted?
Crittora focuses on the execution boundary: whether an AI-enabled system had scoped authority before it called a tool, used an API, or changed a system of record—and what evidence remains afterward.
Direct Answer
What does Fannie Mae Lender Letter LL-2026-04 require?
Effective August 6, 2026, Fannie Mae LL-2026-04 requires covered Single-Family sellers and servicers to maintain policies and procedures governing the development, implementation, use, maintenance, and risk management of artificial intelligence and machine learning used in applicable origination or servicing activity. It also addresses ownership, annual review, information security, vendor and subcontractor governance, and prompt disclosure when Fannie Mae requests information.
What the Requirement Covers
From documented policy to operating evidence
Document the governance program
Maintain transparent policies and procedures for AI and machine-learning development, implementation, use, maintenance, and risk management.
Assign ownership and review
Identify an accountable owner, communicate the program to appropriate personnel, and review it at least annually.
Manage risk and legal obligations
Address trustworthy and ethical characteristics, applicable legal and regulatory requirements, and risk management aligned with organizational tolerance.
Govern vendors and subcontractors
Apply governance to vendor and subcontractor AI or machine-learning use in a manner no less protective than the seller-servicer's own controls.
Connect information security
Align covered AI and machine-learning use with applicable Fannie Mae information-security and business-resiliency requirements.
Prepare to disclose
Be ready to promptly provide requested information about AI or machine-learning types, purpose, manner of use, safeguards, and other requested matters.
AI Governance Questions
Questions About Fannie Mae's 2026 AI Requirements
These answers address Fannie Mae Single-Family Lender Letter LL-2026-04. Fannie Mae Multifamily and Freddie Mac publish separate requirements.
A written policy is necessary but should not be treated as the entire operating program. Fannie Mae LL-2026-04 addresses implementation, use, maintenance, measuring and managing risks, vendor governance, information security, review, and disclosure. Organizations should be able to demonstrate how those responsibilities operate in practice.
#fannie-written-policy-enoughFannie Mae LL-2026-04 applies when a seller or servicer uses AI or machine learning in the origination of mortgage loans sold to or guaranteed by Fannie Mae, subject to the letter's scope and the organization's specific activity.
#fannie-origination-scopeFannie Mae LL-2026-04 applies when a seller or servicer uses AI or machine learning to service mortgage loans on Fannie Mae's behalf, subject to the letter's scope and the organization's specific activity.
#fannie-servicing-scopeYes. Fannie Mae LL-2026-04 expressly addresses both artificial intelligence and machine learning. It is not limited to generative AI, large language models, or public chatbot tools.
#fannie-machine-learning-scopeFannie Mae LL-2026-04 places responsibility on the seller or servicer to ensure that subcontractor and vendor AI or machine-learning use is governed in a manner no less protective than its own program. The organization should identify applicable embedded features, assess risk, document safeguards, and maintain appropriate oversight.
#fannie-vendor-ai-featuresHuman final decision-making does not automatically remove an AI-assisted workflow from Fannie Mae LL-2026-04. The relevant question is whether AI or machine learning is used in covered origination or servicing activity. The workflow's role, data, outputs, human controls, and impact should be documented and assessed.
#fannie-human-final-decisionMISMO's Framework for Responsible AI in the Mortgage Ecosystem provides mortgage-specific resources including governance policy, AI-system inventory, risk-assessment, implementation, and getting-started tools. A lender can use those resources to structure its program while separately mapping its controls to Fannie Mae LL-2026-04. Using FRAME does not by itself certify compliance.
#mismo-frame-and-fannieNo technology product should be described as automatically making a lender compliant with Fannie Mae LL-2026-04. Crittora Agent Authority Broker is designed to support one part of a broader program: execution-time authorization and evidence for sensitive AI actions. Governance, legal analysis, risk assessment, vendor oversight, model controls, security, and accountability remain organizational responsibilities.
#crittora-caab-and-fanniePrimary sources
Use the current official documents when assessing applicability, contractual duties, and implementation decisions.
Crittora provides technology controls and operational information. It does not provide legal advice, certify compliance, or guarantee satisfaction of Fannie Mae, Freddie Mac, MISMO, or other requirements. No endorsement by Fannie Mae, Freddie Mac, or MISMO is implied.
Free Consultation
Schedule a Free AI Readiness Call
Talk with Crittora about your AI-governance priorities, the questions your team is facing, and a practical next step. No lengthy preparation is required.