Book a demo
Topics · GOVERNANCE-AGENTS

FFIEC IT Handbook and Payments Registries | Moring

The FFIEC IT handbook structure gives payment institutions a ready-made frame for agent registries: the IT booklets on audit, information security, wholesale payment systems, and third-party management already describe the controls a payments agent registry must implement, and examiners read agent governance through those booklets today.

On this page
  1. Short answer
  2. How it works
  3. FAQs
  4. Related reading

Short answer

The FFIEC IT handbook structure gives payment institutions a ready-made frame for agent registries: the IT booklets on audit, information security, wholesale payment systems, and third-party management already describe the controls a payments agent registry must implement, and examiners read agent governance through those booklets today.

How it works

The FFIEC IT handbook InfoBase is the examination canon for technology governance at US financial institutions, organized into booklets covering audit, architecture and operations, information security, and payment systems, with definitions, laws and regulations, and references keyed to each booklet. No AI-specific booklet exists yet, which is precisely why payment institutions should read their agent registry against the existing booklets: the controls the handbook describes for systems that touch payment operations apply to agents that touch payment operations. Sources: https://ithandbook.ffiec.gov/ https://learn.microsoft.com/en-us/compliance/regulatory/offering-ffiec-us

Start with the audit booklet's logic. The handbook's audit material covers IT audit roles and responsibilities, independence, internal audit programs, and risk-based auditing, and a payments agent registry fits that frame naturally: the registry is the population against which risk-based sampling proceeds, and its evidence supports the audit program. Institutions should include payment agents in the IT audit universe and test registry completeness, entry quality, and runtime consistency the way they test other inventories, because an incomplete registry is an audit finding in exactly the way an incomplete asset inventory would be, and a payment agent whose recorded authority disagrees with its runtime behavior fails the audit test that matters most. Sources: https://ithandbook.ffiec.gov/

The wholesale payment systems material is the closest booklet analogue. It guides examiners on the risks and risk management practices when originating and transmitting large-value payments, covering operational transaction risk, internal and operational controls, information security, business continuity, and vendor and third-party management within payment operations. A payment agent is a new kind of participant in exactly that risk surface, and the registry operationalizes the booklet's expectations: identify the agent, scope its payment authority, control its access, log its actions, and manage it as a dependency of the payment function. Sources: https://ithandbook.ffiec.gov

The architecture, infrastructure, and operations material contributes the change and lifecycle frame. That booklet's governance sections address board and senior management responsibilities, strategic planning, and enterprise risk management around the technology estate, and a payments registry belongs in that governance: registry standards approved at the management level, lifecycle states for payment agents, and change control that re-approves material scope changes. The twelve-control progression described in agent governance material, from inventory through compliance evidence, reads as a modern restatement of what the booklets have always required, now applied to entities that decide and act. Sources: https://afsaonline.org/wp-content/uploads/2021/07/AIO-IT-Booklet.pdf https://www.moring.ai/topics/payments-control-plane-ffiec

Expect examiners to apply the handbooks to AI without a new rule. Practitioner analysis of AI governance examinations is explicit: although no dedicated AI governance handbook or examination booklet has been issued, institutions should expect examiners to evaluate AI-related risks through existing expectations around governance, information security, third-party risk management, compliance, and model risk management, and the registry is the artifact that organizes those expectations for payment agents. Registry-driven control also gives the audit function the same runtime evidence for agents that it has long had for systems. Sources: https://www.bnncpa.com/resources/preparing-for-ai-governance-examinations-applying-existing-ffiec-expectations-to-emerging-ai-risks https://learn.microsoft.com/en-us/compliance/regulatory/offering-ffiec-us

Build the registry's evidence to booklet standards. Booklet-grade evidence means complete, retained, reconstructable records: agent identity attribution on every write, approvals, decision logs, and the versioned agent definitions behind them, which parallels what the wholesale payment systems material expects for payment operations generally. Because governance is tested at execution, examiners focus on whether boundaries hold when payment workflows run, and a registry whose entries connect to runtime enforcement, tamper-evident logs, and approval records lets the institution answer in booklet vocabulary: controls identified, tested, and operating. Sources: https://www.moring.ai/topics/payments-control-plane-ffiec https://www.bnncpa.com/resources/preparing-for-ai-governance-examinations-applying-existing-ffiec-expectations-to-emerging-ai-risks

The practical takeaway is that no new compliance program is required, only extension. Independent trust-registry initiatives and enterprise agent registries demonstrate the industry converging on inventory, ownership, and trust metadata for agents, while the handbook frame tells payment institutions how examiners will read that metadata: through audit, security, operations, and third-party booklets. Map the registry fields to booklet expectations, put payment agents in the audit universe, and the registry becomes a familiar control rather than a novel one. Sources: https://markets.businessinsider.com/news/stocks/secureauth-opens-industry-first-agent-trust-registry-to-the-public-as-ai-agents-pose-escalating-enterprise-security-threat-1036078378 https://ithandbook.ffiec.gov/ - federalreserve.gov: https://www.federalreserve.gov/supervisionreg/srletters/SR2602.htm - artificialintelligenceact.eu: https://artificialintelligenceact.eu/high-level-summary/ - eur-lex.europa.eu: https://eur-lex.europa.eu/eli/reg/2022/2554/oj

Frequently asked questions

Which FFIEC IT booklets apply to a payments agent registry?

Four in practice: audit, for the registry's role in the risk-based audit universe and evidence; information security, for scoped access and monitoring of payment agents; wholesale payment systems, for the risks of originating and transmitting payments and the controls around them; and the architecture, infrastructure, and operations material, for lifecycle and change governance. No AI booklet exists yet, so examiners read agent governance through these existing booklets, and the registry should be built to satisfy them.

How should payment agents enter the IT audit universe?

As governed entities with a testable inventory. Add the payments agent registry to the audit plan, then test completeness against discovered agents, entry quality against ownership and scope standards, and runtime consistency by sampling entries against actual agent behavior and logs. Risk-based weighting should reflect payment consequence, with consequential-action agents sampled more heavily. An incomplete or stale registry is a finding in exactly the way an incomplete asset inventory would be under the audit booklet's expectations.

What does the wholesale payment systems booklet contribute to agent governance?

The payment-native control frame. It guides examiners on risk management when originating and transmitting large-value payments, covering operational transaction risk, internal controls, information security, business continuity, and vendor and third-party management. A payment agent participates in exactly that risk surface, so the registry operationalizes the booklet: identify the agent, scope its payment authority, control access, log its actions with reconstruction capability, and manage it as a dependency of the payment function with the same rigor as any payment system component.

Will examiners use the IT handbooks against AI agents before new guidance exists?

Yes. Practitioner analysis is explicit that despite no dedicated AI governance booklet, examiners will evaluate AI-related risks through existing expectations around governance, information security, third-party risk management, compliance, and model risk management, and they focus on whether boundaries hold when workflows run. Payment institutions should map registry fields and controls to booklet expectations now, so the registry answers examination questions in familiar vocabulary rather than waiting for AI-specific publication.