Leaving this course?

Lesson 1 of 12

What the relationship actually is

A correspondent relationship is one bank providing banking services to another bank. That sentence contains the whole problem.

The accounts. The respondent bank holds an account with the correspondent so it can make and receive payments in a currency, or in a market, where it has no presence of its own. From the correspondent's books that is a vostro account — "your account with us". From the respondent's own books the same account is a nostro — "our account with you". A small bank in one country can therefore settle US dollar, euro or sterling payments without being licensed in those markets, which is the legitimate and essential purpose of the arrangement.

The customers are not your customers. When the respondent's customer sends a payment, the correspondent executes it — but the person who did the due diligence on that customer is the respondent, not you. You have a customer (the bank) whose business consists of its own customers, whom you cannot see and did not accept. Every difficulty in this field comes from that single fact.

The chain. A payment may pass through several institutions: the originator's bank, its correspondent, possibly a second correspondent, the beneficiary's bank. Each sees part of the picture. A correspondent in the middle may see a payment message, an amount and two names, with no knowledge of the underlying trade, the relationship between the parties, or whether either has been checked properly.

Payable-through accounts. A particular shape worth naming. In some arrangements the respondent's own customers get direct use of the account at the correspondent, operating it as though it were theirs. The correspondent is then exposed to parties who are not its customers and who may not even be known to it. The standards treat this case explicitly because the usual comfort — "the respondent knows its customers" — is doing far more work than it can bear.

Shell banks. A shell bank has no physical presence in the country where it is incorporated and licensed, and is unaffiliated with a regulated financial group. The standard does not ask you to risk-assess these: establishing or continuing a relationship with one is prohibited, and institutions must satisfy themselves that a respondent does not permit its own accounts to be used by shell banks.

Why it matters more than its volume suggests. A single correspondent relationship can carry the payment traffic of an entire bank, and therefore of its whole customer base, across your books. The due diligence you do on one counterparty stands in for due diligence on thousands of people you will never see. That is not a reason to refuse the business; it is the reason it is governed the way it is.

Terminology: the FATF standards say suspicious transaction report (STR), while the United States and the United Kingdom say suspicious activity report (SAR). They are the same filing, and this course uses whichever term its own sources use.

Lesson 2 of 12

What Recommendation 13 requires

Recommendation 13 applies in addition to normal customer due diligence (CDD, which much of the industry calls KYC), not instead of it. It asks for four things.

Understand the respondent. Gather enough information about the respondent institution to understand the nature of its business, and to determine from publicly available information its reputation and the quality of its supervision — including whether it has been subject to a money laundering or terrorist financing investigation or regulatory action. Three distinct enquiries sit in there: what the business is, what others say about it, and how well it is supervised. The third is often the most informative and the most skipped, because it is a question about the respondent's home regulator rather than about the respondent.

Assess its controls. Assess the respondent institution's own anti-money laundering and terrorist financing controls. Not catalogue them — assess them. A completed questionnaire is evidence, not an answer. The industry questionnaire most widely used for this is the Wolfsberg Correspondent Banking Due Diligence Questionnaire, and its value depends entirely on somebody reading the responses against what they know of that market, that size of institution and that regulator.

Senior approval. Obtain approval from senior management before establishing new correspondent relationships. The same logic as any high-risk relationship: a decision by someone accountable, with the file in front of them.

Write down who does what. Clearly understand and document the respective responsibilities of each institution. This is the provision that prevents the most common dispute in a correspondent relationship — each side assuming the other is screening, monitoring or filtering something.

The two specific cases. Payable-through accounts require the correspondent to satisfy itself that the respondent has performed customer due diligence on those customers with direct access, and can provide relevant customer due diligence information on request. The shell bank prohibition is absolute.

What the standard does not say. It does not require you to do due diligence on your respondent's customers. That point is worth stating because it is the misreading that drives withdrawal from the business: an institution that believes it must know every underlying customer will conclude, correctly, that it cannot — and will exit. The requirement is to understand and assess the respondent, and to have the means to ask questions about a specific payment when a reason arises.

Lesson 3 of 12

Nesting, and what a payment message carries

Two technical realities shape what is actually knowable.

Nested correspondent banking. Your respondent may itself provide correspondent services to other banks, so those banks' payments reach you through your respondent's account. This is sometimes called downstream correspondent clearing. The arrangement can be entirely legitimate and is how many smaller institutions reach global markets at all. The risk is specific: you have assessed one institution's controls, and traffic is arriving from institutions you have not assessed, in countries you may not have considered, under the name of the one you did.

Nesting is not forbidden, and it is not the same as a shell bank relationship. What it requires is that you know it is happening — which means asking the question directly, because a respondent that has never been asked may not volunteer it — and that the volume, the countries and the types of institution involved form part of the risk you assessed.

What is in a payment message. Recommendation 16 requires that payments carry required and accurate originator information and required beneficiary information, and that this information stays with the payment through the chain. Two practical consequences:

  • An intermediary institution must ensure all originator and beneficiary information accompanying a transfer is retained with it, and must take reasonable measures to identify transfers that lack required information.
  • Missing or meaningless information in those fields — a name with no address where one is required, a reference where a party should be — is itself a finding, not merely untidy data. The FATF revised Recommendation 16 in June 2025 to reflect modern payment systems and messaging, and the direction of travel is towards more structured, more complete party data rather than less.

Cover payments. In some arrangements the instruction between the two customer-facing banks and the funds movement between correspondents travel separately. Historically the correspondent executing the funds movement could see very little about the underlying parties. Industry standards on payment transparency exist precisely because that opacity was being used, and because an institution that cannot see the parties cannot screen them.

What this means for monitoring. Monitoring a correspondent relationship is not monitoring a customer. You are looking at the shape of the traffic: countries that have no evident connection to the respondent's stated business, a sudden change in volume or value, payments to or from jurisdictions subject to sanctions measures, round-sum round-trip flows, or activity inconsistent with a bank of that size in that market. When something is wrong you usually cannot resolve it yourself; you ask the respondent, under the request-for-information arrangement the relationship should already have provided for.

And sanctions are where this structure bites hardest. A correspondent screens the parties to payments who are nobody's customer — not the originator's bank's, and certainly not its own. It is the point in the chain with the greatest exposure and the least context.

Three things follow. The screening cannot be delegated, whatever the responsibility document says, because the obligation to not make funds available to a designated person attaches to the institution executing the payment. Poor party data is a sanctions problem before it is a data problem: a name with no address or date of birth cannot be meaningfully screened, and a respondent that routinely sends such messages is creating your exposure.

And in several regimes liability does not depend on intent. An institution can breach by processing a payment it had no reason to doubt, which is why the field treats information quality as a control and not as housekeeping.

The deliberate version of the same problem is stripping — a respondent removing or altering party information so that a payment passes a filter. If you find it, it is not a data-quality conversation.

Lesson 4 of 12

De-risking, and what it cost

Correspondent banking is the clearest case study in anti-money laundering of a control causing the harm it was meant to prevent.

What happened. Facing enforcement actions, uncertain expectations and relationships whose revenue did not justify the review cost, large banks withdrew from correspondent relationships, and did so by category: smaller institutions, money transfer businesses, charities, and banks in particular regions. The number of active correspondent relationships worldwide has fallen substantially over the past decade, while the volume of payments has grown — which means the same traffic is concentrated through fewer institutions.

Why that is a problem rather than a tidy-up. Three consequences follow, and they are the opposite of the intent:

  • Less visibility. Payments that previously passed through a regulated correspondent find other routes: longer chains with more intermediaries, or channels outside the regulated system entirely. Nothing stops the money; it stops being observable.
  • Concentration. When a region is served by one or two correspondents, the failure or withdrawal of one is a systemic event for that market, and the remaining institution holds a degree of pricing power over an essential service.
  • Real economic cost. Remittance corridors and trade finance to the affected markets become more expensive and less reliable. The people affected are not the risk the control was aimed at.

The official position. The FATF has been consistent that a risk-based approach does not mean refusing business wholesale, and that de-risking entire classes of customer or country is not what the standards require. Guidance on correspondent banking exists in part to correct the over-reading of Recommendation 13 described in the previous lesson, and bodies concerned with financial stability and development have documented the withdrawal and its effects.

What the alternative looks like in practice. Not a lower standard — a usable one:

  • Assess the respondent, not its customers, and say so in your own policy so staff stop looking for an impossible comfort.
  • Use a standard questionnaire so a smaller respondent is not answering fifty different formats, and read the answers.
  • Tier the book: a respondent in a market with strong supervision, clean traffic and a long record is not the same review as a new relationship in a high-risk market with nesting.
  • Have a working request-for-information channel, with agreed response times, so a specific question can be answered without ending the relationship.
  • Where you do exit, exit for a reason specific to that institution, record it, and give enough notice for its customers to find an alternative.

And if you do exit, the exit is itself a piece of work. "We gave notice" is where most offboarding thinking stops, and it leaves three things undone.

The balance in the vostro account has to go somewhere. A respondent whose account is closed before it has established an alternative cannot move its own customers' money, which is how a compliance decision becomes a liquidity event for another bank.

Payments already in flight will keep arriving for weeks, from institutions downstream that were never told. Decide in advance whether you return them, and to whom.

And the respondent's customers bear it, not the respondent. Importers who cannot pay suppliers, families who cannot receive remittances, a hospital that cannot settle an invoice. None of them chose the relationship and none of them is the risk you were managing.

Notice measured in months rather than weeks, a named contact through the wind-down, and a written reason the respondent can show to a replacement correspondent cost very little and are the difference between an exit and a withdrawal.

Lesson 5 of 12

Running the book

What a well-run correspondent portfolio looks like, and what a supervisor will ask about it.

The file on each respondent should answer, without anyone having to reconstruct it: who owns and controls the institution; what its business is and which markets it serves; who supervises it and how well; whether it has been the subject of regulatory action or investigation; what its own AML programme consists of and who assessed it; whether it offers services to other banks and to whom; whether any of its customers have direct use of the account; and who approved the relationship, when, and on what basis.

Review on risk, not on the calendar. An annual cycle applied to every respondent spends the same effort on the lowest and highest risk relationships. Triggers matter more than dates: a change of ownership, a change in the supervisor's rating of that jurisdiction, enforcement action, a shift in traffic pattern, the discovery of nesting that was not disclosed.

Measure the things that could embarrass you. Not the number of relationships:

  • How many respondents have an overdue review, and for how long.
  • How many requests for information were sent, how many were answered, and how long they took. An unanswered request is a finding about the relationship.
  • How many relationships were found to have undisclosed nesting.
  • How many payments were stopped or returned for missing party information, by respondent — this is the cheapest signal of a respondent with weak controls.
  • How many exits were for a reason specific to the institution, and how many were categorical.

Document who does what, and then test it. The division of responsibility written at onboarding should be checked against what actually happens. If your policy says the respondent screens its customers against sanctions lists and your own screening keeps producing hits on its payments, one of those two things is not true.

The judgement to hold on to. Correspondent banking is infrastructure. A bank in a smaller market that cannot clear dollars cannot serve importers, cannot pay for medicines, cannot receive remittances. The work is to understand the counterparty well enough to carry the relationship deliberately — and to be able to explain, in writing, both the ones you keep and the ones you end.

Lesson 6 of 12

The questionnaire, and how to use it

Due diligence on a respondent bank is largely conducted through a standard questionnaire, and the standard one is the Wolfsberg Correspondent Banking Due Diligence Questionnaire. Its value depends entirely on whether anybody reads the answers.

What it is for. To let one bank ask another, in a consistent form, what its anti-money laundering programme consists of: ownership and licensing, the business it does, which jurisdictions and products, who supervises it, what its policies cover, how it screens, how it monitors, how it trains, and whether it has been subject to enforcement. A common format means a correspondent can compare respondents rather than read fifty different self-descriptions.

The common failure is treating it as a document to obtain. It arrives, it is complete, it is filed, and the relationship is approved. A completed questionnaire is evidence that the respondent has a compliance function capable of filling in a form. The information in it has to be read against what the institution otherwise knows: the payment volumes it actually sees, the corridors it actually observes, the jurisdictions on the file.

Read for inconsistency with the traffic. A respondent describing itself as a domestic retail bank whose traffic through your accounts is overwhelmingly third-country trade payments has told you something, and the questionnaire alone will not. This comparison is the single most useful thing a correspondent can do, and it requires somebody to hold both the answers and the transaction data at once.

Thin answers are answers. "Our policies comply with all applicable laws" in place of a description is a finding. So is a monitoring section that names no scenarios, a screening section that names no list source, and an enforcement section that says none where public records say otherwise.

Ask the questions the form does not. How many staff are in the compliance function, for a bank of that size? Who is the reporting officer, and to whom do they report? How many reports did they make last year — not to judge the number, but to see whether they can answer. When were they last examined and what was found?

Visit, where the relationship justifies it. For a significant respondent in a higher-risk jurisdiction, an on-site visit tells you things no questionnaire will: whether the compliance function exists in more than name, whether the systems described are the systems running, and whether the people doing the work recognise the policy.

Refresh on a cycle and on events. A questionnaire is a snapshot. A change of ownership, a change of supervisor's rating, an enforcement action, entry into a new market, or a material change in the traffic should each trigger a refresh rather than waiting for the anniversary.

And record the assessment, not just the answers. The file should contain somebody's written view: this is what the respondent says, this is what we see, this is what we think, and this is what we will watch. Without that paragraph the questionnaire is paperwork.

Lesson 7 of 12

Monitoring a book you cannot see into

Ordinary transaction monitoring assumes the institution knows its customer and can judge activity against a profile. In correspondent banking the customer is a bank and the activity belongs to that bank's customers, so most of the apparatus does not transfer.

What the correspondent is actually monitoring. Not a person's behaviour but a bank's aggregate traffic. The useful unit is the corridor, the product and the pattern over time, not the individual payment — though individual payments still matter for screening.

Build an expectation, in numbers, and test against it. The approvals file should state what the relationship is expected to do: which corridors, what monthly volume and value, which currencies, what mix of commercial and personal payments, what proportion of cover payments. Without that, the monitoring has nothing to compare against and every alert becomes a conversation about whether the activity is normal.

The signals that are visible at this level. Volume or value stepping to a new level without explanation. A corridor appearing that was never discussed. A currency the respondent's market has no obvious use for. Payments whose originators and beneficiaries are concentrated on a handful of names despite a retail-sounding respondent. Round-amount traffic. Activity peaking just under whatever review threshold the correspondent operates. A sudden change in the proportion of payments where the underlying parties are poorly described.

Originator and beneficiary data is the one window into the respondent's customers. Where the information is complete, it can be screened and patterns in it observed. Where it is routinely incomplete — missing addresses, generic names, a single originator for large volumes — that is itself the finding, and it is one of the few things a correspondent can insist upon contractually.

Watch the shape of the file, not just the transactions. The proportion of traffic that is for the respondent's own account rather than its customers'. The proportion that terminates in the correspondent's jurisdiction against the proportion that passes through. Pass-through traffic is where the correspondent sees least and carries most.

Agree what you will ask, and how fast. A correspondent's main investigative tool is a request for information about an underlying party, and its value depends on the respondent answering properly and quickly. That expectation belongs in the agreement, with a time limit, and the pattern of responses is itself a measure of the relationship: a respondent that takes three weeks to identify its own customer has told you about its systems.

And review the portfolio, not only the relationships. Looking across respondents reveals what looking at one cannot: the same beneficiary appearing through several of them, a corridor growing across the book, two respondents whose traffic is suspiciously complementary. Few institutions do this, and it is where the clearest findings tend to be.

Lesson 8 of 12

Doing something about nesting

Lesson 3 explained what nesting is and why a payment message may not reveal it. This is what a correspondent can actually do about it, which is less than it would like and more than it usually does.

Ask, specifically, and in writing. Does the respondent provide correspondent services to other institutions? Which ones, in which jurisdictions, for what products, and what due diligence does it perform on them? A general assurance that it "applies equivalent standards" is not an answer. The list of downstream institutions is the answer, and a respondent unable to produce one does not know its own book.

Put it in the agreement. The terms can require disclosure of nested relationships, prior notice of new ones, a right to information about underlying institutions, and a right to refuse particular ones. They can prohibit onward provision to institutions in named jurisdictions, or to any institution without notice. These are commercial terms and they are negotiable, which means they have to be asked for.

Then test whether the declaration is true. The traffic will tell you. Payments whose originators cluster around a small number of other banks' customers, originator addresses concentrated in a country the respondent does not operate in, message fields naming institutions that are not the respondent, or volumes out of proportion to the respondent's own size. Undeclared nesting usually shows in the data before anybody admits it.

Treat an undeclared nested relationship as a serious matter. Not because nesting is improper — it is ordinary and necessary, and it is how banks in smaller markets reach the global system — but because the respondent has misdescribed its business to its correspondent, which calls into question everything else it said.

Understand what you are accepting when you accept it. A nested relationship means the correspondent's controls apply to its respondent, the respondent's controls apply to the downstream bank, and the correspondent is relying on a programme it has not seen to manage customers it cannot identify. That reliance is the risk, and it should be recorded in those terms rather than described as "indirect exposure".

Price it, or decline it, but decide it. The options are to permit nesting with disclosure and monitoring, to permit it with restrictions by jurisdiction or product, or to prohibit it. Each is defensible. What is not defensible is a book where nesting is neither permitted nor prohibited and is therefore happening without anybody having decided.

And remember what the alternative was. Prohibiting nesting outright across a book removes access for exactly the institutions least able to obtain a direct relationship, which is the de-risking harm of the previous lesson arriving by another route. The proportionate answer is usually disclosure, information rights and attention to the traffic, not a blanket refusal.

Lesson 9 of 12

Sanctions in the middle of a payment chain

A correspondent sits between parties it does not know, moving money it did not originate, in a chain where it may be neither the first nor the last institution. That position creates the sanctions exposure the industry has been penalised for most heavily.

The screening question is different here. The correspondent screens message fields rather than customers: originator, beneficiary, both banks, intermediaries, and the free-text fields where information about the purpose or the underlying parties may sit. Everything depends on what the message contains, which is why data quality in payment messages is a sanctions matter and not only an operational one.

Incomplete data defeats the control, and defeats it silently. An originator recorded as a name with no address, or a beneficiary described generically, cannot be screened to any useful standard. The alert does not fire and nothing indicates that the control did not work. This is why a correspondent is entitled to insist on complete party information and why accepting degraded data is a decision with a consequence.

Stripping is the conduct that produced the largest penalties. Removing or altering references to sanctioned parties or jurisdictions from payment messages so that they pass screening downstream. It is deliberate, it has been penalised very heavily across the industry, and it is worth naming plainly in training: altering a payment message to defeat another institution's controls is not a technicality.

Cover payments and the information gap. Where a payment is sent by one route and its cover by another, the institution receiving one part may not see the underlying parties at all. Industry standards changed to address this, and the correspondent's practical question is whether it can see the underlying parties on the traffic it handles, and what it does when it cannot.

Jurisdictional reach is wider than people expect. Clearing a currency generally brings the clearing jurisdiction's rules into the transaction regardless of where the parties are. A payment between two parties in countries with no restrictions can still be prohibited because of the currency it settles in, and respondents in other markets frequently do not appreciate this. Explaining it is part of running the relationship.

Hold rather than release when unresolved. A payment is irreversible once made. The one thing a correspondent should never do under time pressure is release a payment with an unresolved possible match, which means the authority to hold must sit with the person on shift and not two levels above them.

And tell the respondent what you expect. Complete party data, no alteration of messages, prior notice of business in restricted markets, and an undertaking to answer information requests quickly. All of it belongs in the agreement, because a correspondent's sanctions exposure is created by its respondents' behaviour and governed by its own contracts.

Lesson 10 of 12

The things that are simply prohibited

Most of this course is judgement. A small part is not, and the hard rules are worth separating out because they are absolute and because institutions occasionally discover they have breached one.

Shell banks. Recommendation 13 prohibits entering into or continuing a correspondent relationship with a shell bank, and requires institutions to satisfy themselves that a respondent does not permit its accounts to be used by shell banks. The definition is the point: a bank incorporated in a jurisdiction where it has no physical presence and which is unaffiliated with a regulated financial group. Physical presence means meaningful management and administration actually located there, not a registered agent and a mailbox.

Which makes the question practical, not formal. Where are the people? Who makes the credit decisions? Where are the records kept? A licence and an address prove nothing by themselves; a jurisdiction with a large number of licensed banks and very few bank buildings should prompt the question.

And the second limb is the one that gets missed. It is not enough that your respondent is not a shell bank. You must be satisfied that it does not let shell banks use its accounts — which is the nesting question from the previous lesson wearing its most serious hat. A respondent that cannot name its downstream institutions cannot give you that assurance, whatever it writes on a form.

Payable-through accounts. Where a respondent's own customers get direct access to the correspondent account and transact on it themselves, the standards require the correspondent to be satisfied that the respondent has identified and performed due diligence on those customers and can supply the relevant information on request. In practice this is a high bar, and many institutions simply do not offer such arrangements, which is a defensible position.

Senior management approval, and what it means. Recommendation 13 requires approval from senior management before establishing a new correspondent relationship. As with the politically exposed person requirement, the test of whether it is real is whether it has ever produced a refusal, and whether the approver was given the weaknesses rather than a recommendation.

Respective responsibilities, written down. The standards expect the two institutions to understand and document who does what. In practice this means the agreement should say what the respondent will do about its own customers, what information it will provide and how quickly, what it will disclose about nesting, and what it will not do with payment messages.

None of this is risk-based. The prohibitions apply regardless of how attractive the business is, how long the relationship has run, or how well the respondent scores on every other measure. That is unusual in this field and worth stating in training, because staff trained to weigh everything will try to weigh these too.

Lesson 11 of 12

Exiting without doing harm

Lesson 4 described what wholesale withdrawal cost. An institution will still sometimes need to end a relationship, and the difference between a managed exit and de-risking is mostly in how it is done.

Separate the two decisions. Whether to exit this respondent, on these facts, is a risk decision. Whether to exit a category — a country, a sector, a size of bank — is a strategy decision with consequences beyond the institution. Conflating them is how a single bad case becomes a withdrawal from a market.

Before exiting, ask what the alternative looks like. Restriction is usually available: closing particular corridors, prohibiting nesting, excluding a product, capping volume, requiring additional information, moving to a shorter review cycle. A relationship reduced to what the institution can actually see and manage is often better for everybody than the same relationship ended.

If the answer is exit, give real notice. A respondent that loses its only route to a currency with thirty days' notice may be unable to serve its own customers, including remittance recipients and importers who have no part in whatever caused the decision. Notice periods long enough to find an alternative are a cost to the exiting institution and the main thing that distinguishes a managed exit from an abrupt one.

Phase it where the traffic is significant. Stopping new business while allowing existing commitments to run down, in agreed stages, avoids stranding payments in flight and gives the respondent something to plan around.

Say why, as far as you lawfully can. Where the reason is a specific, fixable deficiency, telling the respondent gives it the chance to fix it — and the sector the chance to improve. Where the reason is a suspicion that has been reported, the restrictions on disclosure apply and the institution should say that it cannot give reasons rather than inventing a commercial one.

Tell the respondent's supervisor where that is appropriate. A bank losing correspondent access is a matter its own regulator will want to know about, and in some jurisdictions there are channels for exactly this. An institution that exits silently leaves the authority to find out from the consequences.

Record the decision as a decision. What was considered, what alternatives were examined and rejected, who decided, and over what period. This is the record that answers the supervisory question about de-risking, which is no longer whether the institution exited but whether it did so case by case with reasons.

And keep count. How many respondents the institution has exited, from which jurisdictions, for what reasons, over what period. An institution that cannot produce that table cannot know whether it has been making individual decisions or quietly leaving a region.

Lesson 12 of 12

One worked example, end to end

A respondent of nine years. A mid-sized commercial bank in a developing market, holding a dollar account used for trade settlement and remittance payouts. Monthly volumes have been stable for years. The annual review falls due.

What the questionnaire said. Domestic commercial bank, trade finance and retail, no correspondent services provided to other institutions, policies in place, monitoring in place, no enforcement history. Complete, signed, unremarkable — and in the previous four years, filed.

What the traffic said. The reviewer did what nobody had done before: put the answers beside twelve months of payment data. Volume had grown 40% in eighteen months, within the limits nobody had set because the approval file recorded no expected volume at all. More usefully, the originators had changed shape: where the traffic had once been spread across hundreds of names, a third of the value in the last six months came from payments whose originator addresses clustered in a neighbouring country the respondent does not operate in.

The question that followed. Not "is this suspicious" but "whose customers are these". A request went to the respondent for the originator detail on twenty payments. The answer took nineteen days and identified, for eleven of them, another bank.

So the respondent was nesting, and had said it was not. It transpired that two institutions in the neighbouring market had lost their own direct relationships two years earlier and had been served through the respondent since. The respondent's view was that it had not misled anybody because it regarded these as commercial customers rather than correspondent relationships, which is a distinction its own questionnaire did not permit.

What the correspondent decided, and what it rejected. It did not exit. The downstream banks were licensed and supervised, the underlying payments were ordinary trade and remittance flows, and ending the relationship would have cut off two further markets. It restricted instead: nesting permitted but only for the two named institutions, each disclosed and documented, no further downstream institutions without prior notice, complete originator data required, information requests answered within five business days, and a six-month review cycle with the volume expectation now written down.

The conditions were given owners and dates, and the agreement was amended to carry the disclosure and information terms it had never contained.

What the institution learned about itself. Four things went into the follow-up log. No approval file in the correspondent book recorded an expected volume, so growth could not be an alert. Nobody had ever compared a questionnaire against traffic, which is the cheapest control available here. No agreement in the book required disclosure of nesting. And a portfolio-level review had never been run, which would have shown the same originator cluster arriving through two other respondents as well.

Knowledge check

Fifty questions

Answer all fifty questions, then check your answers. You need 45 out of 50 to pass and receive a certificate. If you score less, you will see which answers were right and wrong, and then go through the course again before you retake the check. Your answers, progress and times are kept only in this browser.

Standards and frameworks covered

The published frameworks this course teaches toward. Saolix is not affiliated with, endorsed by or accredited by any of these bodies; naming a standard is a citation, not a claim about its author.

Sources

The official documents this course relies on. Laws and guidance change, so check the current version.

  1. Guidance on Correspondent Banking Services (2016) · Financial Action Task Force
  2. The FATF Recommendations (Recommendations 13, 14 and 16, and their Interpretive Notes) · Financial Action Task Force
  3. Revision of Recommendation 16 on payment transparency (June 2025) · Financial Action Task Force
  4. FATF Guidance on the Risk-Based Approach and De-risking · Financial Action Task Force
  5. Guidance for a Risk-Based Approach for Money or Value Transfer Services · Financial Action Task Force
  6. Country Risk Frequently Asked Questions (2024) · The Wolfsberg Group
  7. Frequently Asked Questions on Risk Assessments for Money Laundering, Sanctions and Bribery · The Wolfsberg Group
  8. Trade Finance Principles (2019) · Wolfsberg Group, ICC and BAFT
  9. Guidance on Sanctions Screening (2019) · The Wolfsberg Group
  10. Guidance for a Risk-Based Approach for the Banking Sector · Financial Action Task Force