Advisory

AI Governance

Financial services firms are putting AI into onboarding, fraud detection, compliance, and customer-facing products. Regulators have taken notice. Examiners now have frameworks for reviewing it: state regulators through the CSBS AI Supervisory Framework, and the OCC through its Super Intelligence Supervisory Framework for national banks. Both treat AI as an extension of the model risk management, governance, and compliance programs a firm already runs, which means examiners now have a structured way to ask how those programs account for it.  

FS Vector helps banks, fintechs, and the companies that sell AI to them build the governance, testing, and documentation that make those deployments defensible to examiners, bank partners, and customers.

What Is AI Governance?

AI governance is the set of policies, controls, and accountability a firm uses to decide where AI is used, how it is tested, and who answers for its output. It builds on the model risk management, third-party risk, and compliance programs a firm already has rather than standing apart from them. The OCC's framework is explicit on this point: AI is not a separate risk category, and the AI-specific considerations it describes supplement existing practices. Both regulators have acknowledged the gap this leaves: the CSBS examiner guide notes that existing model risk resources may not fully address generative or agentic AI, and the OCC has said the model risk principles in its revised model risk management guidance (Bulletin 2026-13) apply to non-generative, non-agentic models. Those considerations exist because generative and agentic AI strain the assumptions traditional validation relies on. Their output varies from run to run, their failure modes are harder to predict, and they often come from third-party vendors that can change the model between reviews. Extending an existing program to cover those risks is real work: new use case inventories, new risk classifications, new testing methods, and new vendor terms.

How Does FS Vector Approach AI Governance?

We treat AI the way examiners and bank partners will: as part of the compliance program. The same view runs through all our practices: examination management, embedded banking, due diligence, and independent assessment each now include AI in scope, and each considers good AI governance.

Whether deploying AI tools for the first time or enhancing an existing program, we help clients put the right controls in place and show that they work:

  • AI governance programs (policies, oversight structures, roles, and accountability for AI used in AML/KYC, fraud, onboarding, and underwriting)
  • Independent AI risk reviews (expert review and structured scenario testing of AI tools, including consumer-facing GenAI, for compliance, consumer harm, UDAP/UDAAP, fair lending, and accuracy risks)
  • AI vendor due diligence (helping banks and firms evaluate AI vendors at onboarding and through ongoing oversight)
  • Examiner and partner documentation (clear explanations of how an AI system works, how it is controlled, and how issues are escalated)

For Companies Selling AI to Banks

For a company that provides AI tools to banks or fintechs, the buyer's diligence process is the sales cycle. The OCC's framework describes the AI-specific questions a bank may add to the third-party diligence it already performs under interagency guidance: the scope of authority and data access for a provider's agents, an incident vocabulary and notification timelines, contractual responsibility for model errors and losses, and notice before AI features are added or changed.  

We support fintechs and technology providers through bank partnership diligence every day, and for AI providers we prepare the governance documentation, testing evidence, and model descriptions that bank third-party risk teams ask for.

Ready to Get Started? Connect With Our AI Governance Team

Thanks! A member of our team will reach out within 1 business day.
Oops! Something went wrong while submitting the form.

AI Governance: Featured Team Members

View all

AI Governance FAQs

Does an existing model risk management program cover generative AI?

Partly. It remains the foundation, and regulators treat AI governance as an extension of it, not a separate program. But traditional model risk management was designed for models with defined inputs, outputs, and performance metrics. Generative AI produces variable output, is often supplied by a vendor, and can fail in ways that standard validation does not test for. We help firms extend their governance to cover those risks without rebuilding the program from scratch.

Is a firm still responsible for a vendor's AI product it uses?

Regulators and bank partners generally expect firms to understand and oversee the third-party tools they rely on, especially where those tools touch customers or compliance decisions. Our vendor due diligence work helps firms document that oversight and identify gaps before an examiner or partner does.

What does an independent AI risk review involve?

We start with an expert review of the tool's design, controls, and intended use against the compliance and consumer protection risks most relevant to the client's business. Where needed, we follow with structured scenario testing, including edge cases, to see how the tool behaves in practice. The client receives a gap assessment and prioritized recommendations.

Can FS Vector help an AI company selling to banks?

Yes. We support fintechs and technology providers through the bank partnership process every day. We help AI vendors anticipate bank diligence questions and prepare the governance documentation partners expect to see.

How is this different from a technology audit?

Our focus is regulatory and consumer risk, not code quality or infrastructure. Our team of lawyers, advisors, and compliance practitioners reviews AI the way an examiner or bank partner would.