BrokerVA
BrokerVA Team · November 3, 2026

What State Examiners May Ask About AI, and About Your Vendors

What State Examiners May Ask About AI, and About Your Vendors

On September 16, 2026, the state regulators' association CSBS published a framework showing what examiners may ask about AI, including AI embedded in your vendors' software. It creates no new legal obligations, and each state decides whether to use it. Here is what it covers and how to prepare, as part of our guide to compliance for outsourced operations.

What Is the CSBS AI Supervisory Framework?

A set of tools for state examiners. The Conference of State Bank Supervisors released its AI Supervisory Framework for examiners of state-chartered banks and state-licensed nonbank financial companies, which includes licensed mortgage companies. According to American Banker's reporting, it has five documents: a core examiner guide, a 28-page work program, a supplement for nonbanks, a risk-tiering worksheet, and a list of source materials.

Two facts matter most. First, it is discretionary: the core guide says it "does not create new legal obligations or supervisory requirements." Second, each state agency decides how much of it to adopt, and CSBS publishes no count of which states have. So whether your examiner uses it depends on your state. CSBS says it published the framework partly so institutions can prepare for examinations.

What Does an Examiner Ask First?

Eight questions, as American Banker summarized the core guide:

  1. Does the institution use AI at all?
  2. Has it identified where?
  3. Does AI touch customers or shape decisions?
  4. Does it rely on vendors or outside platforms for AI?
  5. Has it looked for AI embedded in vendor products it already uses?
  6. Does it use generative AI?
  7. Does it sort its AI uses by risk?
  8. Does customer or other sensitive information pass through any AI?

Notice questions 4 and 5. If the answer to the first question is unclear or negative, the guide says an examiner may check it against your vendor inventory, software inventory, approved tools, and recent product changes, because AI may be missed when it is built into software as a feature.

What Documents Might They Request?

Records that show you know where AI is and who answers for it. Coverage of the framework lists:

  • AI policies and procedures, and inventories of AI systems and use cases
  • Governance and oversight materials, and board reporting
  • Risk assessments, and monitoring and control records
  • Samples of AI-generated material that customers see, such as chatbot transcripts
  • Vendor documentation and contracts
  • Contract terms covering how AI systems use, retain, share, or train on customer, confidential, or institutional data

See Venable's summary and American Banker's report.

What Does It Say About Vendors?

That vendor AI deserves its own look. The Nonbank AI Supplements add AI-specific considerations to existing reviews of third-party and vendor oversight, model risk, and consumer protection. According to White & Case, the vendor supplement helps examiners ask AI-specific questions within existing vendor oversight reviews, and a vendor-provided AI underwriting tool may raise third-party, model-risk, and consumer-protection issues at once, as one law firm summary notes.

Another firm notes that the framework points examiners to AI-specific due diligence, contractual protections, vendor monitoring, change management, and contingency planning. Many financial companies will not build their own AI, so vendors are where much of your AI sits.

How Does Risk Tiering Work?

Through an optional worksheet that rates each AI use. American Banker reports it scores four factors: how much the use affects consumers, how much human oversight it gets, how much harm an error or outage could cause, and how sensitive the data is. Each use lands in one of three tiers, and controls accumulate as risk rises:

  • Tier 1: an inventory entry naming a business owner, and a written acceptable-use policy.
  • Tier 2: documentation of how the AI reaches outputs customers see.
  • Tier 3: all of that, plus independent model validation and incident response procedures written for AI.

CSBS did not set dollar or volume thresholds. It asks that expectations scale to an institution's size, complexity, risk profile, and use of AI, and the same questions apply to institutions of all sizes.

How Does It Fit With the Other Rules?

It sits alongside them. Freddie Mac's Guide Section 1302.8 has required an AI governance framework since March 3, 2026, and Fannie Mae's Lender Letter LL-2026-04 took effect August 6. National Mortgage Professional's summaries say both reach vendors' AI. The FTC's Safeguards Rule expects you to oversee your service providers. See Who Is Accountable When AI Touches a Mortgage File? and What Should Your Outsourcing Contract Say About a Data Incident?

Survey data suggests where gaps may show. In the AARMR, MBA, and BCG survey, 87% had written AI policies but 58% reported ongoing monitoring, and in Ncontracts' 2026 survey, 72% of organizations were only partly aware of which vendors use AI. One is small and the other vendor-run, so treat them as indicators.

How Do You Prepare?

These steps are our recommendations, and not legal advice.

  1. Build an AI inventory, including AI embedded in software you already use, and approved and unapproved tools.
  2. Ask your vendors where AI is used, and for what. A vendor can add AI features after you sign.
  3. Write an acceptable-use policy, and name an owner for each use.
  4. Tier your uses, using the worksheet even though it is optional, and keep it as support.
  5. Check your contracts for terms on whether the vendor uses, keeps, shares, or trains AI on your data.
  6. Define the human checkpoints, and keep records of who reviewed what. See How to Add AI Without Losing the Human Check.
  7. Add AI to your incident response plan, and keep an evidence file so a request doesn't start from scratch.
  8. Watch your state regulator for whether and how it adopts the framework.

What Should You Ask Vendors About AI?

In writing, and the same questions of each:

  • Do you use AI in work on our files or data? Where, and for what?
  • Is any of our data used to train or improve AI, by you or a third party?
  • Who reviews AI-assisted output before it affects a file, and how is that recorded?
  • Can you describe your AI use in a form we could give an examiner?
  • Will you tell us before you add AI features that touch our data?
  • How do your subcontractors use AI?

See Security Due Diligence for Outsourced Mortgage Operations and What the 2026 Vendor-Risk Survey Means for Outsourced Mortgage Operations.

How Does BrokerVA Fit?

BrokerVA provides mortgage operations support through 150+ specialists in five stage-based roles, Setup, Disclosure, Closing, Funding, and Post-Closing, and accountability for the loan stays with you and your licensed people. We have partnered with Addy AI for LOA services. We haven't published a full statement of our AI use on this page, and we'd rather you ask than assume. Expect to put the questions above to any provider, including us, and we'll talk them through on a call. We don't claim to make anyone compliant.

Our goal is to lower your cost to originate. Every client gets a dedicated Coach for a 6-week implementation, who learns your processes, workflows, and expectations, then ongoing coaching and regular check-ins. We can't promise a specific saving, since it depends on your operation. On a call, we'll walk through your numbers and show you our cost calculator live. Pricing is customized to your volume and the roles you need. See how BrokerVA works and State Examiners and Remote Team Supervision.

Frequently Asked Questions

What is the CSBS AI Supervisory Framework? A set of examiner tools released September 16, 2026, for state-chartered banks and state-licensed nonbanks: a core guide, a work program, nonbank supplements, a risk-tiering worksheet, and a source list.

Does it cover vendors' AI? Yes. The nonbank supplement addresses third-party and vendor oversight, and examiners are asked whether an institution has looked for AI embedded in vendor products.

What should a mortgage company do now? Build an AI and vendor inventory, ask vendors where AI is used, check contracts for data-use terms, tier your uses, define human checkpoints, and keep an evidence file.

Sources

Doing a vendor review that includes AI? Book a call with BrokerVA and bring your questions. We'll walk through your numbers and show you our cost calculator.

BrokerVA | NMLS #1977844