What Financial Sanctions Look Like From the Inside
By Heorhi Tratsiak
The message came back at 11:47 on a Thursday morning.
Not dramatically. Not with any alert that distinguished it from the hundred other messages processed that morning. A SWIFT reply, formatted correctly, arriving through the standard channel. MT199 — a free-format message, the network’s equivalent of a handwritten note slipped under a door when the formal language runs out.
The content, translated from the careful institutional language of international banking correspondence, said: we require additional supporting documentation before this transaction can be processed.
The transfer itself was unremarkable. A customer credit transaction. A woman in Berlin sending money to her mother in Minsk. The sort of transfer I had processed thousands of times. The correspondent bank had cleared identical transactions from the same originating institution, in the same category, for years. Without comment.
I looked at the documentation package. It was complete. Name, account number, declared purpose, supporting identification. Nothing in it that differed from any prior transaction of the same type.
I requested the additional documentation from the client anyway, because that was the procedure, assembled the revised package, and sent it to the correspondent bank the same afternoon.
The reply came back three days later.
More questions.
I had seen this pattern once before, briefly, in 2020, when the first tremors ran through our institutional relationships following the events of that summer. I recognised its texture. But in 2020, the questions had eventually stopped. The relationship had stabilised. The friction had subsided.
This time was different. I understood that within the first week. The questions were not going to stop. They were going to multiply. And behind the questions — I understood this too, though no one said it directly — was a decision that had already been made, not about this transaction, not about this client, but about our institution, our jurisdiction, and what category of risk we now represented to every counterparty in the global financial network.
The correspondent account had not gone silent in a single moment. That is not how it works. It had begun, quietly and methodically, to go deaf.
This is the story of what happened next.
The Woman in Berlin
Her name, for the purposes of this account, is Alina.
She is thirty-four. She has been living in Germany for three years, working in a distribution warehouse on the outskirts of Berlin. Her shift starts at six in the morning. She sends money home to her mother every month — always on the same day, the twenty-eighth, which is when the rent comes due in Minsk. Her mother is sixty-seven, with a pension that does not fully cover her costs, and a phone that she struggles to operate, and a habit of standing at the kitchen window at midday on the twenty-eighth, watching the notification that does not come.
Alina is not a financial person. She does not think about SWIFT. She does not know what a correspondent account is, or what MT103 means, or that the euros she earns at the warehouse in Berlin must pass through a chain of institutions — three, sometimes four — before they become Belarusian rubles in her mother’s account. She knows only that she sends the money and it arrives. She has been sending it for three years and it has always arrived.
On the morning of February 28, 2022, it did not arrive.
Alina checked her German bank’s app. The transfer had been initiated. The amount had left her account. It showed as pending, then as processing, then, after two days, as a status her German bank’s app described only as under review.
She called the bank’s helpline. She waited thirty-one minutes. The representative told her that the transfer was experiencing a delay and that she should allow additional processing time.
She called again the next day. She was told the same thing, in slightly different words.
Her mother waited by the window. The rent came due. She borrowed from a neighbour, which she had not done since her husband died fifteen years ago, and which cost her something that had nothing to do with money.
Alina did not know what had happened. The transfer was eventually returned to her account, seven weeks later, with a reference code that meant nothing to her. The accompanying message, in dense institutional language, said that the transaction could not be processed due to compliance review requirements under applicable regulatory frameworks.
She tried again through a different service. That transfer also failed.
She eventually found a workaround — a partial solution involving cash, a third country, and two acquaintances willing to act as intermediaries. It was expensive, slow, and undignified. She used it every month for the rest of that year.
Her mother’s neighbour was repaid. But the debt — the specific, invisible debt that belongs to a sixty-seven-year-old woman who has had to ask for help because a system she did not know existed decided, without knowing her name, that she was an acceptable cost — that debt was never settled. Systems do not settle those kinds of debts. They do not know how.
I did not know Alina. I did not process her specific transfer. But I processed transfers like hers, every day, for years. And in February 2022, I sat at my terminal and watched transfers like hers start to fail, one by one, while I tried to understand exactly what was happening and found, for the first time in my career, that the system I had spent seven years learning was no longer behaving as I had been taught.
What Happens Before the Silence
To understand the silence, you have to understand what the sound was like before it.
In a functioning correspondent banking relationship, the rhythm is unremarkable. Instructions leave your terminal in standardised SWIFT format. They arrive at the correspondent. They are processed. The confirmation comes back — typically within hours for routine transactions, a day at most for anything requiring additional review. The settlement happens. The obligation moves. The ledger updates.
You do not think about this process any more than you think about breathing. It is the background condition of the work. Its normalcy is so complete that you only notice it the way you notice air — when it becomes difficult to find.
The deterioration of a correspondent relationship does not announce itself. It arrives as a gradual increase in the cost of doing ordinary things.
First, the documentation requests become slightly more thorough. A transaction that had previously required standard identification now requires additional information about the purpose of the funds. Not unreasonable. Compliance departments update their requirements. You provide the documentation.
Then the processing times lengthen. A transfer that cleared in twenty-four hours now takes forty-eight. Then seventy-two. The correspondent bank’s automated replies are unchanged in tone. Everything is routine. There are simply additional steps in the review process.
Then the questions become more specific. A compliance contact — someone you have perhaps spoken to twice in three years — sends a personal message asking about a particular category of transactions. Nothing formal. A professional courtesy. An early-warning signal from someone who, in my experience, genuinely did not want the relationship to deteriorate and was giving you the opportunity to address it before it became official.
This is the friction layer. I want to be precise about what this term means, because it describes a mechanism that almost no account of financial sanctions has adequately captured.
The friction layer is not prohibition. It is not the formal, legally defined moment of disconnection from SWIFT or the official designation of an institution as a sanctioned entity. It is the accumulation of compliance costs, documentation burdens, processing delays, and relationship deterioration that precedes any formal action — and that, in most cases, does the majority of the work that formal sanctions are credited with.
An institution operating under heavy friction does not need to be officially sanctioned to be effectively isolated. The practical effect — the inability to move money efficiently, the erosion of counterparty confidence, the progressive unavailability of services — can be largely achieved through friction alone. The formal designation is, in many cases, the final administrative act in a process that has been running for months.
I watched the friction layer activate in the weeks that followed the geopolitical escalation of late February 2022. I watched it from inside an institution that was not itself the primary target of the sanctions regime but that existed, in the eyes of every correspondent bank’s compliance function, in a jurisdiction and a corporate network that had become, overnight, a category of elevated risk.
We were close enough to the perimeter that the edge effects were operationally real. And the edge effects were this: every correspondent relationship we maintained became, simultaneously, a question.
February 24, 2022
The invasion began on a Thursday.
I do not need to describe what that morning felt like in the country where I worked, because I am not writing a political account. I am writing an operational one. And operationally, the first hours were characterized by something that is, in retrospect, almost strange: the systems kept working.
The SWIFT terminals processed the morning’s instructions. The card networks authorised transactions. The correspondent banks were open for business. The machinery of the global financial system, for the space of a few hours on February 24, 2022, was continuing to turn as if nothing had changed.
This is the temporal gap that nobody describes, because nobody outside a financial institution is positioned to observe it. The geopolitical event occurs. The world changes. The financial infrastructure — built for stability, designed to be slow to fail — takes time to register the change and respond.
The response, when it came, did not come as a single event. That is the most important thing to understand about how financial sanctions work in practice, and it is the thing that every analysis written from outside an institution systematically misrepresents.
There was no switch. There was no morning when we arrived at our desks to find that a single decisive action had changed everything. What there was, instead, was a sequence. A cascade. A thousand individual compliance decisions, made by a thousand different institutions in a dozen different countries, each acting rationally in its own interest, each producing an incremental reduction in the functionality of a system that the previous day had seemed entirely stable.
This is what I have called, in the preceding chapter, the compliance cascade. But I want to define it here precisely, because the precision matters.
The compliance cascade is the mechanism by which a regulatory signal at the centre of the global financial network — a sanctions designation, an OFAC listing, a regulatory guidance — propagates outward through the compliance decisions of private institutions at the network’s periphery, producing field effects that extend far beyond the formally designated targets, through the rational, self-interested behaviour of institutions that are not themselves subject to the regulation but are subject to its consequences.
No institution in this cascade makes a decision to harm anyone. Each institution makes a decision to protect itself. The cumulative effect of those protective decisions is the systematic isolation of everything within a certain radius of the designated entity. The radius is not defined anywhere in any legal document. It is determined by the risk calculations of compliance departments in banks you have never heard of, in cities you have never visited, making judgments about categories of exposure that include your institution as a data point in a spreadsheet.
This is not cynicism. It is the system working exactly as designed. The designers, however, were not thinking about Alina’s mother waiting for her rent money. They were thinking about something much larger. The individuals caught in the blast radius were, in the formal calculus of the policy, acceptable costs.
I want you to sit with that phrase for a moment. Acceptable costs. It is the language of policy analysis, where real people become variables and suffering becomes a coefficient in an equation whose output is measured in geopolitical pressure. The phrase is not malicious. It is something worse than malicious. It is accurate.
What the Terminal Showed
I want to describe what the following weeks looked like from the position I occupied, because the operational detail carries something that the analytical frame alone cannot.
In the first days following the geopolitical escalation, the most visible change was not in the volume of failed transactions but in the volume of questions. Emails from compliance contacts at correspondent banks asking for clarification on specific transaction categories. Requests for updated documentation on our institutional profile — the questionnaires that correspondent banks use to assess their exposure to risk, updated annually under normal conditions, now arriving mid-cycle. Requests for information about specific clients and their transaction histories.
Each question was individually answerable. We answered each one, with the documentation they asked for, within the timeframes they specified. This was the professional reflex of an institution that had spent years maintaining its relationships with care.
What became apparent within the first two weeks was that the questions were not, in any meaningful sense, looking for answers. They were a form of institutional behavior that serves a different function: the documentation of due diligence. When a correspondent bank eventually decides to reduce or exit a relationship, it needs to show its own compliance function and its own regulators that it conducted appropriate review before doing so. The questions are the evidence of that review. The answers to the questions are, in a sense, secondary.
I understood this gradually, through the accumulation of signals that individually admitted other explanations. A compliance contact who had always responded within twenty-four hours now took three days. A category of transactions that had processed automatically was rerouted to manual review. A particular corridor — transfers to specific destination accounts — began generating systematic requests for additional originator information even when the originator information was already on file.
The terminal showed all of this. It showed the MT199 replies accumulating. It showed the processing times extending. It showed the ratio of approved transactions to transactions-under-review shifting in a direction that, once you understood its meaning, did not shift back.
What the terminal could not show — what no system in any bank I am aware of was designed to capture — was the human cost of each failed or delayed transaction. Behind every transfer that stalled in the correspondent queue was a person who needed the money to move and did not understand why it wasn’t moving. A person who had, in many cases, already told someone on the other end to expect it.
The Clients We Couldn’t Help
During those weeks, I had conversations with clients that I have not been able to fully leave behind.
A man who ran a small import business. He had a payment due to a supplier in Germany. The payment was legitimate, the relationship established, the documentation complete. The transfer failed twice. On the third attempt, it entered a review process from which it did not emerge for eleven days. His supplier, operating with their own cash flow constraints, sent an increasingly formal sequence of emails. The business relationship, which had taken two years to build, sustained damage in eleven days that took the better part of a year to partially repair.
He came to me not in anger, exactly, but in the particular frustration of someone who has been failed by a system they thought they understood. He wanted me to explain what was happening. I gave him the most honest answer I could, which was technically accurate and practically useless: international payment processing was experiencing delays due to changes in the compliance environment. The situation was being monitored. Additional documentation requests were being responded to as quickly as possible.
What I could not tell him was what I actually knew. I could not tell him that the correspondent bank’s compliance function had reclassified our institution’s jurisdiction as elevated risk, and that this reclassification had moved all transactions from that corridor into a manual review queue that was processing at approximately one-third of its normal speed because the volume of reviews had tripled overnight while the staffing had not changed. I could not tell him that this situation was not a temporary anomaly but a structural shift that was likely to persist and might deepen. I could not tell him that his supplier’s emails, legitimate and reasonable as they were, were in a causal chain that had started not with any failure of our institution but with a decision made in the Treasury Department of a country neither of us would ever visit, by people who had never heard of his business.
I could not tell him these things for the same reason that the system could not help him: because the truth, in this case, was not actionable. Knowing the mechanism does not stop the mechanism.
A woman with a daughter studying in Poland. Monthly transfers to cover tuition and living expenses. The first two months after February 2022 processed with delays. The third month, the transfer failed entirely. She came to the branch carrying a printout of her daughter’s tuition invoice and a look I recognised from the crisis years of 2020 — the look of someone who has done everything correctly and is trying to understand why that isn’t enough.
Her daughter was twenty-one. She was studying economics, which I found, in the circumstances, too precise to be ironic and too ironic to be precise. She was studying the theory of financial systems while the practical reality of those systems was preventing her tuition payment from arriving. She eventually received a grace period from the university. The transfer eventually cleared, through a route that required additional steps her mother had not previously needed and could not fully explain to her daughter over the phone.
These are not dramatic stories. That is the point. They are the texture of what financial infrastructure weaponisation looks like at the human scale — not the geopolitical analysis, not the macro-economic impact, but the specific, ordinary disruption of specific, ordinary lives. The woman who cannot explain to her daughter why the money is late. The businessman whose two-year relationship sustains damage in eleven days. The sixty-seven-year-old woman in Minsk standing at a kitchen window, watching for a notification that does not come.
The policy was not designed for these people. They were not considered. They were the acceptable costs of a calculation made at a scale where individual human experience is not a visible unit of measurement.
The Cascade Reaches the Consumer
Below the institutional layer, something else was happening simultaneously — something more visible, more personal, and in certain ways more immediately devastating.
The card networks moved faster than the correspondent banks. Within days of the sanctions announcements, Visa and Mastercard had suspended operations for cards issued by designated Russian institutions. The consumer card layer of the financial system — the layer that most people interact with every day without ever thinking about its architecture — simply stopped working for millions of people who woke up one morning to find that the plastic in their wallets had become inert.
This is worth understanding at the operational level, because the speed at which it happened was itself extraordinary. The correspondent banking tightening I have been describing took weeks to reach its full effect. The card network suspension happened in forty-eight hours.
The reason for the difference is architectural. Card networks are centralised in a way that correspondent banking is not. Visa and Mastercard are single entities with the ability to flip a switch at the network level. When they decide to suspend services to cards issued by a particular institution, the effect propagates globally, instantaneously, through their own infrastructure. Every terminal in every country that runs on Visa’s network checks authorisations against Visa’s central systems. When those systems are updated to decline cards from affected issuers, every terminal in every country declines those cards.
The humans who experienced this were in airports, in foreign hospitals, in supermarkets in countries they had moved to for work. They were trying to pay for groceries. They were trying to pay for prescriptions. They were standing at ATMs in cities where they had no alternative access to cash, watching a decline message appear where their balance should have been.
Some of them had time to prepare. Many did not. Financial disruption does not schedule itself conveniently. It arrives on the morning you have a payment due, a flight to catch, a medication to collect.
I was not on the card operations side of the business during this period. But I knew the architecture from my earlier work on card network certifications, and I understood what was happening in operational terms. The network was executing decisions that had been made at a policy level, through mechanisms that had been built for exactly this purpose and that functioned, in a technical sense, perfectly. The systems worked as designed. What they were designed to do, it turned out, included this.
There is a particular kind of helplessness that comes from watching a technically functional system produce human damage. The machine is not broken. The machine is doing exactly what the machine does. The suffering is not a malfunction. It is an output.
What the Silence Sounds Like
By April of 2022, the texture of our correspondent relationships had changed in ways that were, by then, measurable.
I will not give specific numbers, because the specific numbers are not mine to publish. What I can describe is the quality of the change, which is more instructive than any particular statistic.
Certain transaction corridors that had operated smoothly for years were now processing at a fraction of their previous efficiency. Certain correspondent relationships that had been active and functionally normal were now, effectively, in a managed wind-down — not formally closed, not officially terminated, but responding with a slowness and a caution that communicated, to anyone who had been in this business long enough to read its signals, that the relationship was ending.
The formal language was always the same. Compliance review requirements. Additional documentation needed. Processing delays due to heightened due diligence standards. The formal language did not say: we have decided that the regulatory risk of maintaining this relationship exceeds its commercial value. It did not need to say that. The conclusion was the same regardless of the language used to arrive at it.
What this meant in practice was that certain things became impossible. Not prohibited. Impossible in the operational sense — not because any rule said they could not happen, but because the infrastructure through which they would need to happen had progressively degraded to the point where the practical outcome was indistinguishable from prohibition.
This is the deepest mechanism of the financial infrastructure as a weapon, and it is the one most rarely described. The formal sanctions — the SWIFT disconnections, the official asset freezes, the explicit designations — are the visible tip of a much larger structure. Beneath them, working through the compliance behaviour of thousands of private institutions, is a distributed system for making things progressively, invisibly harder. Not impossible. Just harder. And then harder still. And then harder again. Until the cumulative difficulty of doing normal business in a normal way crosses some threshold and the outcome of normal business disappears.
The genius of this mechanism — if genius is the right word for something whose consequences I am describing in terms of a sixty-seven-year-old woman borrowing from her neighbour — is that it requires no central decision-making. Nobody issues an order. Nobody takes responsibility for the specific harm to any specific person. Each institution in the cascade makes a rational, defensible, locally reasonable decision. The aggregate outcome of those locally reasonable decisions is a global effect that, if you described it as a policy goal and someone asked who had authorised it, would be difficult to assign to any single actor.
This is the distributed architecture of financial coercion. Power without clear ownership. Effect without clear cause. Suffering without clear author.
The Question Nobody Could Answer
In the months that followed, I had a question that I carried with me and could not resolve.
The question was not about whether financial sanctions were justified or unjustified in the specific geopolitical context that had generated them. That is a question for historians, ethicists, and policymakers, and I am none of those things. I am a person who processed international payment transfers and worked with correspondent banks and certifed for card network services, and who watched, from that position, what happens when the infrastructure of the global financial system is used as a coercive tool.
My question was more specific. It was: who, in this chain, is accountable for Alina’s mother?
Not legally accountable. Not in any procedural sense. I mean accountable in the older, plainer meaning of the word: who can give an account. Who can explain, in a way that honours the specificity of her situation, why she was standing at that kitchen window, waiting for money that was hers, that had been sent to her, that her daughter had earned through legitimate work, and that did not arrive because a correspondent bank in a third country had updated a risk classification in a system that her daughter had never heard of and would not have understood if someone had explained it.
The sanctions policy exists at a level of abstraction where that question cannot be heard. The compliance decisions that implement it exist at a level of procedural specificity where that question is outside the scope. The financial institutions at every node in the chain were each following their own rules, which were each reasonable, which were each individually defensible.
Nobody made a decision to harm Alina’s mother. The harm was the output of a system that was not designed to account for Alina’s mother, because it was not designed at the level of individual people at all.
This is not an argument against sanctions. It is not an argument for them. It is an observation about what they are, at the level of operational reality, that the policy debates that produce them almost never include.
The people who design financial sanctions think about Iran’s oil exports, about Russia’s central bank reserves, about the economic incentive structures of governments and elites. They think at the level of macroeconomic impact, of leverage, of strategic outcome. These are the relevant variables at the level at which the policy operates.
Alina’s mother is not a variable. She is a person. And the financial infrastructure weapon, in its extraordinary, efficient, distributed power, makes no distinction.
The Weight That Does Not Leave
I left the institution in the autumn of 2022. I have written about that decision elsewhere, and I will not repeat its full account here.
What I want to say here is something different. It is about what stays with you after you have worked inside a system during the period when it reveals what it actually is.
I spent seven years learning the machinery of global finance. I learned it with genuine interest and professional care. I understood the SWIFT message formats, the correspondent banking mechanics, the card network certification processes. I understood how money moves and why it sometimes doesn’t. I was good at this work, and I respected it.
What I did not fully understand, until I watched it happen, is that every piece of knowledge I had accumulated was also, in the right circumstances, the knowledge of how to operate something that causes harm. Not through any decision I made. Not through any failure of professional conduct. Simply through the logic of the system itself: that the infrastructure which connects people and enables their commerce is the same infrastructure that can be used to disconnect them and disable their commerce. That the skills required to operate the first are identical to the skills required to implement the second. That there is no version of this technical knowledge that is available only for the benign application.
This is the weight that does not leave.
It is not guilt. I did not make the policies. I did not freeze the reserves or suspend the card services or update the risk classifications. I processed the transactions as correctly as the constraints allowed, and when the constraints made correct processing impossible, I told clients the truth — or as much of the truth as was available and useful to tell.
But I was there. I was part of the infrastructure. My work, on the days when the system functioned as designed, contributed to the functioning of a system that, on other days, used that same functioning to produce outcomes I would not have designed and cannot defend.
There is no clean resolution to this. The people who work inside financial systems do not get to choose which version of the system they are operating on any given day. They work within the structure as it exists. The structure, as it exists, is capable of both.
Alina found her workaround. Her mother received her rent money, eventually, through a route that was more expensive and more complicated and required the involvement of people who should not have needed to be involved. The neighbour was repaid.
The mechanism that produced this disruption is still fully operational. It has not been reformed or reconsidered. It has, if anything, been further developed and refined. The next time a geopolitical decision requires the financial infrastructure to contract around a target, it will do so with greater precision and speed than the last time.
The people who will be caught in the next compliance cascade — the next Alina, the next mother waiting at the kitchen window — have no way to prepare for it, no recourse within it, and no mechanism to hold accountable the distributed system that produced their harm.
They are acceptable costs in a calculation that has no line for them.
I think about them often. I will think about them for a long time.
Because I was inside the machine when it ran. Because I know what it is made of. Because I know, with the particular precision that comes from operational knowledge, exactly how it works.
And because knowing how something works is not the same as knowing how to stop it.
What This Chapter Means for Anyone Working in Finance
I want to end with something practical, because this book is not a lament. It is an account, and an account serves a purpose only if it changes how people think.
If you work in financial services — in payments, in compliance, in correspondent banking, in card network operations — you work inside the infrastructure I have been describing. You work inside a system that is, simultaneously, the most powerful engine of human economic cooperation in history and the most powerful non-military instrument of coercion ever built.
These are not two different systems. They are the same system. The features that make it possible for Alina to send money to her mother across a border are the same features that made it possible, in 2022, to prevent that money from moving. The correspondent relationships you maintain in your institution are the same relationships that, under the right regulatory signal, become the mechanism of a compliance cascade.
This is not a reason to leave the industry. It is a reason to understand it.
The operational person is not the policymaker. You do not decide which countries are sanctioned, which institutions are designated, which transactions are prohibited. What you do decide — what remains, irreducibly, in your domain — is the quality of attention you bring to the people on the other side of every transaction you process. Whether you treat each transfer as a number in a queue or as the movement of an obligation between two human beings who are counting on it to arrive.
This distinction does not change the policy. It does not protect Alina’s mother. But it is the only part of the system that is fully within your control, and in a machine this large, that is not nothing.
The silence of a correspondent account is absolute from the inside. The client sitting across from you hears a different silence: the silence of a system that was supposed to help them and has, without explanation, stopped.
What you can give them — the only thing, sometimes, that is left to give — is honesty about what you know and care about what it costs them.
I gave that when I could. I regret the times, which were more frequent toward the end, when the institutional constraints on what could be said made even that impossible.
The stone does not lift. But some things deserve to be named for what they are.
This is one of them.
