Tornado Cash and the transfers nobody asked for

The week

Most sanctions stories have a villain with a name. This one had a piece of software.

On August 8, 2022, the Treasury Department’s Office of Foreign Assets Control, or OFAC, added Tornado Cash to its Specially Designated Nationals list. OFAC is the office that decides who Americans may not do business with. Its list is the government’s master register of blocked people, companies and, increasingly, things. Once something is on it, U.S. persons cannot transact with it, and any property in which it has an interest must be frozen.

The legal hook was Executive Order 13694, the authority for sanctions on malicious cyber activity. The action was not the first against a mixer. In May 2022, OFAC had designated Blender.io, a smaller Bitcoin mixer, also citing North Korean laundering. Nor was it the first time digital currency addresses had appeared on the list. OFAC had been adding them to individual entries since late 2018.

Tornado Cash was a virtual currency “mixer” on the Ethereum blockchain. A blockchain is a shared public ledger that records every transfer of a digital currency, visible to anyone. A mixer pools many people’s deposits and pays them back out, so that an outside observer can no longer trace which coins went where. A mixer works like a coat check where everyone hands in identical black coats and later collects one at random. You leave with a coat, but nobody can say whose it was. That is useful for people who want ordinary financial privacy. It is also very useful for thieves.

The Lazarus Group was already well known to sanctions teams. In April 2022, the FBI had attributed the March theft from the Ronin bridge, the system behind the online game Axie Infinity, to Lazarus. OFAC added an Ethereum address used in that theft to the group’s existing entry on its list. Bridges, which move assets between blockchains, had become a favored target because they hold large pools of funds controlled by code.

Treasury’s announcement said Tornado Cash had been used to launder more than 7 billion dollars’ worth of virtual currency since 2019. That included more than 455 million dollars stolen by the Lazarus Group, a North Korean state-sponsored hacking group. Treasury also said more than 96 million dollars from the June 2022 Harmony bridge theft, and at least 7.8 million dollars from the Nomad bridge theft of August 2, had passed through it. The release told virtual currency firms that mixers should generally be treated as high-risk. It also recalled that FinCEN had assessed a 60 million dollar civil penalty in 2020 against the operator of a different mixer under the Bank Secrecy Act. And it pointed firms to OFAC’s October 2021 sanctions compliance guidance for the virtual currency industry, which had set out expectations for screening wallet addresses and using geolocation tools.

What made the designation unusual was what it listed. It named dozens of Ethereum addresses, many of them “smart contracts.” A smart contract is a small program stored on the blockchain that runs automatically when someone sends it funds, with nobody able to unplug it. These contracts ran autonomously, and the designation barred U.S. persons from transacting with them.

The consequences arrived within days. Circle, the issuer of the USDC stablecoin, froze the USDC held at the sanctioned addresses, reported at roughly 75,000 dollars. GitHub, where developers store and share code, suspended accounts linked to the project, including those of its developers. On August 10, Dutch authorities arrested Alexey Pertsev, one of its developers, in Amsterdam on suspicion of money laundering.

Meanwhile an anonymous user sent small amounts of ether, Ethereum’s currency, through the mixer to the public wallets of well-known people, a practice called dusting. Those recipients had done nothing, yet they now held funds that had passed through a sanctioned address. Some decentralized finance services began blocking wallets that had received the dust from their websites, catching the unwilling recipients. Some infrastructure providers that relay transactions to the Ethereum network began refusing requests involving the listed addresses. Compliance by intermediaries, not by the code, was how the designation took effect. Critics began arguing that immutable software, code that cannot be changed or switched off, could not lawfully be sanctioned at all. The first lawsuit followed in early September.

The banking agencies were also busy with crypto that month. On August 16, the Federal Reserve told the state member banks it supervises to notify it before starting crypto-asset activities. The FDIC and the OCC already had similar notification expectations in place. Supervisors wanted to know which banks were entering crypto before they did, and sanctions controls were part of what they would ask about.

For banks, credit unions and fintechs, the week raised a practical question: how do you comply with a sanction when there is no counterparty to refuse?

Take a compliance officer at a mid-sized bank on the morning of August 9, 2022. Her sanctions program, like every bank’s, is built around counterparties. A customer, a beneficiary or an intermediary bank has a name and an identity, and the institution decides whether to deal with them. Her screening software compares names on payments against the government’s list and stops the matches. The process is mature and well understood.

The Tornado Cash designation removed both of the anchors that process depends on. The sanctioned thing was largely code that nobody could switch off. And anyone could use it to send funds to anyone else, including to people who had never asked for them.

The designation raised a new set of questions for banks, credit unions and fintechs with any crypto exposure, directly or through partners. They fall into three parts: funds that arrive without being requested, exposure several transactions away, and how to run a control while its legal basis is contested. Before those, the machinery. After them, what the episode means for institutions with no crypto business at all, and where the case against this approach is strongest.

Ledgers, mixers and code nobody can switch off

Public ledgers and private keys

Start with the blockchain. Ethereum, like Bitcoin, is a ledger that thousands of computers around the world keep in sync. Every transfer is written down permanently and publicly. Nobody needs a bank account to use it; you need a “wallet,” which is essentially a pair of digital keys. The public half is your address, a long string of letters and numbers that works like an account number anyone can see and send money to. The private half is your password, the thing that lets you spend what is held at that address.

This openness is the first surprise for many executives. Crypto is often described as anonymous, but on a public blockchain every transfer between addresses is visible to anyone who cares to look. What is hidden is the identity behind each address. If you can link an address to a person, for example because they withdrew money through an exchange that checked their passport, you can often trace a great deal of their activity.

A second distinction matters for compliance. Some wallets are “custodial”: an exchange or other firm holds the private keys on the customer’s behalf, much as a bank holds deposits. Others are “self-hosted”: the individual holds the keys and no institution stands in between. A custodial firm can stop a customer from moving funds. Nobody can stop the holder of a self-hosted wallet, except by refusing to deal with the addresses they use.

How the mixer broke the trail

That traceability is precisely what a mixer is designed to defeat. Tornado Cash ran separate pools for fixed amounts of ether, such as 0.1, 1, 10 and 100. A user deposited one of those amounts and received a secret note. Later, often much later, the user presented a cryptographic proof that they held a valid note, and the contract paid the same amount to a fresh address.

The proof used a technique called a zero-knowledge proof. It lets someone show that a statement is true without revealing why. Here it proved “I made one of the deposits in this pool” without revealing which one. Because every deposit in a pool was the same size, the amounts gave nothing away either. The coat-check analogy holds: the attendant never wrote down whose coat was whose.

One practical wrinkle remained. Sending a transaction on Ethereum costs a small fee, paid in ether. A brand-new address has no ether to pay it. Tornado Cash allowed independent “relayers” to submit withdrawals on the user’s behalf and take their fee from the withdrawn amount. That kept the fresh address free of any link to the user’s other wallets.

Code without an operator

Tornado Cash ran this pool through smart contracts. Traditional services have an operator who can be served with a court order or told to freeze an account. A smart contract has no such person in the loop. Once deployed, the program lives on the blockchain and executes for anyone who interacts with it, according to its code, indefinitely. Many of the Tornado Cash contracts could not be altered or shut down even by the people who wrote them. The developers had deliberately given up that control, which was part of the project’s pitch to privacy-minded users.

That is a different situation from a stablecoin like USDC. A stablecoin is itself a smart contract, but its issuer wrote in the ability to freeze balances at particular addresses. When Circle froze the USDC at Tornado Cash addresses, it was using a switch it had built in. Ether has no such switch. Nobody, including the Ethereum developers, can freeze ether at an address.

What blocking means

Finally, the sanctions rule itself. When OFAC designates a party, U.S. persons must not deal with it. If they come into possession of property in which a blocked party has an interest, they must block, or freeze, that property and report it to OFAC. A bank that receives a wire for a sanctioned company does not send it back; it holds the money in a blocked account and tells the government.

The reporting has a set rhythm. OFAC’s reporting rules require an initial report within ten business days of blocking. Holders of blocked property also file an annual report of what they hold as of June 30 each year. Civil liability for sanctions violations is strict, meaning a violation can be penalized even without intent. OFAC’s enforcement guidelines weigh factors such as whether the institution had a compliance program and whether it disclosed the problem voluntarily.

These duties apply to all “U.S. persons.” That covers every U.S. bank, credit union, broker-dealer, investment adviser, fintech and crypto exchange, and their staff. It also covers foreign branches of U.S. banks. The obligation is not limited to the institutions that the Bank Secrecy Act calls financial institutions, although examiners usually review sanctions compliance alongside anti-money-laundering programs.

Those duties were written for a world of named counterparties. In August 2022 they were applied to a program with no owner.

Funds nobody requested

The duty does not depend on consent

The dusting episode made the first problem concrete. Anyone holding a public wallet address could receive funds from a sanctioned contract without doing anything. Under the rules as written, a U.S. person who receives property in which a blocked party has an interest must block it and report it to OFAC. A transfer that arrives unasked does not change that obligation.

In September 2022, OFAC issued answers to frequently asked questions. It acknowledged that the rules technically reached unsolicited dust. It said it would not prioritize enforcement against delayed blocking reports from people who had received such dust with no other sanctions connection. It also said that people who had deposited funds before the designation could apply for a license to complete their withdrawal, and that it would view such requests favorably. And it said that copying, discussing or teaching the open-source code was not, by itself, prohibited.

That eased the pressure on individuals. It left institutions with an operational task.

Why the freeze is the whole control

Crypto parts company with traditional banking here. On a blockchain, the platform cannot reject an inbound transfer the way a bank can reject a wire. There is no “return to sender” button; the funds simply appear. The platform can only decide what to do with the funds after they land, which makes the freeze and reporting steps the whole control.

Sending the funds back is not an option either. Returning them would itself be a transaction with a blocked party. So the only lawful path is to hold them, record them and report them.

For a custodial exchange, holding is straightforward in principle. The firm can move the specific amount to a segregated wallet or flag it in its internal ledger, so the customer cannot spend it. For a self-hosted wallet held by an individual, “blocking” mostly means not touching the funds and reporting them. Institutions should get their own counsel’s view on how to treat dusted balances that are mixed with legitimate funds at the same address.

The written procedure

A crypto platform or a bank’s crypto partner needed a written procedure for inbound exposure. It should say how a receipt from a sanctioned address is detected and who decides whether to freeze the affected balance. It should also cover how the report is filed and how the customer is told without disclosing more than the rules allow.

Consider a crypto exchange that serves retail customers through a partnership with a regional bank. One Tuesday, a small amount of ether lands in a customer’s deposit address straight from a listed Tornado Cash contract. A well-run exchange has already decided, in writing, that its screening tool will flag the receipt within minutes. A named sanctions officer reviews it the same day. Only the specific amount received is ring-fenced, so the rest of the balance stays usable where the rules allow. The blocking report goes to OFAC on time, and the customer receives a short, approved message.

A poorly run exchange discovers the receipt weeks later during an audit. It freezes the whole account in a panic and leaves the contact center to improvise. Both firms had the same exposure. Only one had a defensible record.

Where it lands differently

The same dust plays out differently by type of institution.

• A custodial crypto exchange registered with FinCEN as a money services business holds the keys, so it can and must act on the funds directly. Its sanctions program is the front line.
• A bank offering crypto custody is in the same position for custodied assets. It also has to fit the dusted balance into its blocked-property reporting and its examiner’s expectations.
• A credit union working with a third-party digital asset provider, as NCUA guidance from December 2021 allowed federal credit unions to do, usually does not hold the crypto at all. Its exposure is through the provider, and its control is the contract and the oversight.
• A community bank that banks crypto firms, or holds deposits for a fintech program, rarely sees on-chain activity. It sees dollars arriving from an exchange after a customer sells. Its exposure is indirect and depends on how well the exchange screened.
• A payments fintech or neobank that lets customers buy crypto through an embedded provider stands between the two. It owns the customer relationship but often not the keys, and its sponsor bank inherits whatever gaps the arrangement leaves.
• A broker-dealer in 2022 typically had little direct crypto activity, but customers moving money to and from exchanges still touched it.

The common thread is that the institution closest to the keys carries the operational burden. Everyone further away carries the oversight burden.

Exposure several hops away

What a hop is

The second problem is distance. A bank’s direct dealings with a mixer were usually nil. Its exposure ran through three kinds of relationship: business customers that are crypto firms, fintech programs where the bank holds customer funds, and retail customers who move money to and from exchanges.

Screening an address list for exact matches is easy. The meaningful question is whether funds reaching a customer passed through a sanctioned address recently, and how many intermediate transfers make a difference. Each transfer between addresses is called a “hop.” Money that went from the mixer straight to your customer is one hop away. Money that went from the mixer to a stranger, then to an exchange, then to your customer is three hops away.

Analytics and its judgment calls

Answering the question requires blockchain analytics: specialist software that maps the public ledger, labels known addresses and traces how funds flowed between them. The labels cover exchanges, mixers, sanctioned parties, known thieves and much else. It also requires a written policy on hop distance and risk scoring. A policy set too tight flags most of the market, because so much crypto has brushed past something unsavory at some point. One set too loose misses obvious laundering chains.

The analytics involve modeling choices that most buyers never see. Tainted and clean funds are often combined at one address and then split, and the tool has to decide how much of each outgoing payment counts as tainted. One approach treats every output as fully tainted. Another assigns taint in proportion to what went in. A third follows the order in which funds arrived. Each gives a different answer for the same transaction, and the choice drives how many alerts an analyst sees.

Attribution is the other weak point. Labels on addresses come from the vendor’s own investigations, public sources and inferences about which addresses belong together. They can be wrong, and they can be missing. New York’s financial regulator made a related point in April 2022. Its guidance to the virtual currency firms it licenses said blockchain analytics should support due diligence, monitoring and sanctions screening, and it acknowledged the limits of attribution.

Where the work falls

For most banks, this work falls to partners. The practical first step was written confirmation that crypto business customers and fintech partners screen on-chain activity. Banks also needed to understand the analytics provider and thresholds used, and to obtain the right to review alerts and dispositions. Crypto business customers warranted enhanced due diligence on their sanctions programs, including how they handle dusted funds. Retail monitoring needed rules that catch transfers from platforms known to process mixer outflows without treating every crypto-related payment as suspicious.

Many banks also treat screening tools as models for governance purposes. That brings them under model risk management: documented assumptions, independent validation and ongoing performance monitoring. Applying that discipline to an analytics vendor’s risk scores is harder than applying it to a name-matching engine. The vendor’s methods are proprietary, and the ground truth about who owns an address is often unknown.

How on-chain screening runs day to day

In a traditional bank, sanctions screening happens at a few fixed moments. A customer is onboarded, a payment is sent or received, or the government updates its list. A name goes in, the software looks for close matches, and an analyst reviews anything flagged. Most alerts are false positives, people who happen to share a name with someone on the list. The analyst clears them with a note explaining why.

On-chain screening adds a second layer. When a crypto firm, or a bank’s partner, receives or sends funds, the analytics tool looks at the address on the other side. It also looks at where that address’s money came from. It returns a risk score and a short explanation, something like “a meaningful share of these funds passed through a sanctioned mixer within the last few transfers.” An analyst then decides whether that is close enough to matter under the firm’s written policy.

Three operational realities follow. First, the address lists change. When OFAC adds addresses, the analytics provider has to load them quickly, and the firm needs to know how quickly, because the obligation applies from the moment of designation. Second, the work never finishes. Because the blockchain is permanent and public, funds that looked clean last month can look different once new addresses are labeled. Firms need a view on when to look back at past activity. Third, the decisions need a paper trail. An analyst who clears an alert should record why, in language a regulator or an auditor could follow a year later.

Designations also raise a look-back question. When OFAC adds addresses, a firm should decide whether to screen its historical transactions against them. Dealings that took place before a designation were not sanctions violations. They can still reveal customers who deserve a closer look, a changed risk rating or a suspicious activity report. A written rule on how far back to look, and for which designations, keeps the decision consistent rather than ad hoc.

For a bank that relies on partners, all of this happens out of sight by default. It is done inside the partner’s systems and by its staff. So the right to see alerts and decisions, rather than a summary assurance, matters a great deal.

Contract terms worth having

Partner agreements signed before 2022 rarely anticipated on-chain sanctions exposure. A bank renegotiating or renewing one would usually want several specific terms:

• A description of the partner’s on-chain screening, including the analytics provider, and notice before either changes.
• A commitment to load new OFAC address designations within a stated time.
• Prompt notice to the bank of any blocking event involving funds that touch the bank’s accounts or customers.
• Access to alert data and dispositions for the bank’s customers, in a form the bank can test.
• Audit rights covering the sanctions program, exercisable by the bank or its independent auditor.
• Retention of screening records for the period the bank’s own policy requires.

Partners may resist some of these, particularly sharing data from a proprietary analytics tool. The negotiation itself is informative. A partner that cannot describe its own thresholds is telling the bank something about its program.

Metrics an examiner would ask for

A bank overseeing crypto exposure through partners should be able to produce a small set of evidence on request:

• The written hop-distance and scoring policy, with the date it was approved and by whom.
• The time between an OFAC list update and the analytics tool reflecting it, measured, not assumed.
• Alert volumes, how many were escalated, and how long reviews took.
• A sample of cleared alerts with the analyst’s rationale.
• Results of testing the policy against known laundering chains, including what it missed.
• Records of any blocked property and the related reports.

The cost does not scale down

Boards needed to hear one implication plainly. Offering digital asset services, or banking those who do, carries a fixed cost in analytics tools and specialist staff that does not scale down with volume. A bank with ten crypto business customers needs most of the same capability as one with a thousand. Whether the business line earns its keep is a strategic question as much as a compliance line item.

Supervisors were also watching entry into the business closely. In 2022, national banks needed to show the OCC they could conduct crypto activities safely before starting. FDIC-supervised banks were asked to notify the FDIC. The Federal Reserve added its own notification expectation for state member banks the week after the Tornado Cash designation. For a bank weighing a crypto line in that environment, the sanctions capability was part of what it had to demonstrate before it could begin.

A control under legal attack

The challenges

The third problem was legal. Critics argued that an ownerless smart contract is not property that can be blocked under the International Emergency Economic Powers Act, the statute behind most sanctions programs. They also argued that the designation swept in lawful users seeking financial privacy. A group of Tornado Cash users, with financial backing from Coinbase, sued Treasury in federal court in Texas in September 2022. The policy group Coin Center filed a separate challenge the following month.

In November 2022, OFAC removed and simultaneously re-listed Tornado Cash. The new designation added North Korea sanctions authority alongside the cyber authority. OFAC also clarified that it had not designated the project’s founders, developers, governance participants or users. The underlying obligation did not change.

Comply while the case is pending

The challenge did not suspend the obligation in 2022. The designation was in force, and U.S. persons had to comply. A bank that decided on its own that the designation was probably unlawful, and stopped screening for it, would have had no defense if the courts disagreed.

Running a control whose basis may fall calls for documentation. Each control should record which designation it implements, the risk rationale, and what would change if the designation were narrowed or withdrawn. That makes a later change a configuration update rather than a program rebuild. It also preserves the record of why balances were frozen, which matters when a frozen customer asks for release.

Plan the exit as well as the entry

The same documentation should cover the exit. If a designation is lifted, blocked balances do not release themselves. Someone has to confirm the change, check that no other sanctions connection applies to each account, release funds in an orderly way and close the related reports. A procedure written in advance avoids a scramble and inconsistent treatment of customers in the same position.

Delisting also does not erase other obligations. Funds that passed through a mixer may still warrant suspicious activity review under anti-money-laundering rules, even when no sanctions block applies. A firm that releases a balance because the sanctions basis is gone should still ask whether the activity behind it needs a report.

What the board should hear

A contested designation puts two risks in front of a board at once. One is the compliance risk of failing to block. The other is the customer and reputational risk of blocking funds that a court may later say should never have been frozen. The first risk carries legal penalties; the second carries complaints and, possibly, litigation from customers.

The sensible framing for directors is that the institution follows the law as it stands and records its reasoning. It does not take a position on whether the designation is wise. Management should report on the volume and value of blocked crypto-related property, the number of customers affected and the status of the legal challenges. That gives the board a basis for judging whether the exposure is worth the business it supports.

Legal monitoring is part of the control. Someone should own tracking the litigation, OFAC’s FAQs and any amended designations, and should trigger a review of procedures when any of them changes. In this case the designation itself was amended within three months, and the FAQs arrived within five weeks.

Scripts for the front line

Frontline staff needed scripts as well. Two different confidentiality regimes apply, and staff should know which is which. Telling a customer that funds have been blocked under sanctions rules is generally expected; the customer needs to know why they cannot use the money. Disclosing that a suspicious activity report has been filed, or is being considered, is prohibited by law. A script that blurs the two can cause a violation in either direction. Customers whose transfers were held often called the branch or contact center. A consistent explanation that the institution must follow federal sanctions rules, without discussing specific screening results, protected both the customer relationship and the institution. How sanctions obligations apply to a particular institution’s activities, and to balances it has already frozen, is a question for its lawyers.

A rule for people, applied to software

Why it matters without a crypto business

The episode matters even to a bank with no crypto business. Tornado Cash was an early, vivid example of a regulator applying a rule designed for people and companies to an automated system that has no owner in the usual sense. The questions it raised will recur wherever money moves through software rather than through institutions.

The stakes cut across the usual executive concerns. The money at risk on any single dusting transfer was trivial, but the regulatory exposure was not. Sanctions violations are judged strictly, and an institution that cannot show it had a process can face penalties and lasting supervisory attention. Customers are affected directly when balances are frozen, and how that is handled shapes trust. Staff in the contact center bear the brunt of confused or angry callers. And reputation is on the line both ways: being seen to process North Korean laundering proceeds is damaging, but so is freezing ordinary customers clumsily.

Where the pattern recurs

The Tornado Cash designation also exposed a gap that runs through decentralized finance more broadly. Bridges, decentralized exchanges and lending protocols are all built from smart contracts. Some have administrators who can pause or upgrade them; others do not. Treasury’s own release named two bridge thefts from that summer, and both of those bridges were, in effect, code holding pooled funds.

For a bank, the question each of these raises is the same. Where the counterparty is code, the control has to sit with the regulated institutions that touch the funds before and after the code. Exchanges, custodians, stablecoin issuers and the banks behind them are those institutions. The more of the chain that runs through software, the more the remaining human checkpoints carry.

Probabilistic controls need a stated appetite

There is also a governance point. Sanctions decisions in traditional banking are mostly binary: the name matches or it does not. On-chain screening is probabilistic. A risk score says “this money is somewhat associated with something bad.” Probabilistic controls need explicit risk appetite, set by senior people and written down, or they default to whatever the vendor’s settings happen to be.

Writing that appetite down forces choices that are otherwise left implicit. The number of hops that matters for sanctions purposes may differ from the number that matters for anti-money-laundering monitoring. The share of tainted funds that triggers review has to be set somewhere. The answer may differ for a retail customer and a business customer. Those are policy decisions, and the people accountable for them should be identifiable.

The counter-arguments

The strongest case against this approach comes from three directions.

The first is privacy. Ordinary people used Tornado Cash to keep their salaries, donations or holdings from public view on a ledger anyone can read. Treating every brush with a mixer as suspect penalizes legitimate privacy. That argument has force, and the courts later accepted part of the legal version of it. It does not change what a U.S. institution had to do while the designation stood.

The second is de-risking. A bank that responds by refusing all crypto-related customers avoids the problem by pushing it elsewhere. That can cut off legitimate businesses and push activity to firms with weaker controls. The better answer is a calibrated policy, which costs more to run than a blanket ban.

The third is that analytics are too unreliable to support decisions about customers. Attribution can be wrong, and taint models disagree. That is a reason to document how the scores are used and to keep a human decision in the loop. It is not a reason to ignore on-chain evidence that is often clearer than anything a wire transfer reveals.

What the stronger programs shared

A strong program in this area tends to share a few features.

• Clear ownership. One named person or team owns on-chain sanctions exposure, including what partners do on the bank’s behalf.
• Written thresholds. The hop-distance and risk-scoring policy is documented, approved, and tested against known laundering chains, so the bank knows what it would and would not catch.
• Contractual visibility. Partner agreements give the bank the right to see alerts, decisions and the analytics provider’s settings, rather than a promise that screening happens.
• An entry and an exit. Procedures cover both freezing funds when a designation appears and releasing them when one is lifted.
• Prepared people. Frontline staff have an approved script, and customers hear the same explanation whichever branch they call.

Weaker programs got the same things wrong. Some institutions assumed that because they did not touch crypto directly, they had no exposure, and never asked their fintech partners. Others bought an analytics tool and accepted its default settings without anyone deciding whether those settings matched the bank’s risk appetite. Some treated every crypto-related transfer as suspicious, which drowns analysts in alerts and alienates ordinary customers. And many had no plan for what would happen if the legal ground shifted. They were left to reconstruct, months later, why particular balances had been frozen.

The cost of the stronger approach is real. It needs specialist staff who understand both sanctions law and blockchain data, which were scarce in 2022. It needs contract negotiations with partners who may resist sharing alert data. And it needs senior time to set and revisit thresholds. For an institution whose crypto exposure is small, the honest choice may be to limit that exposure deliberately rather than build a full capability.

Where crypto touches the bank, and who decides what to freeze

• Where, specifically, does crypto touch us: business customers, fintech partners, retail customers moving money to exchanges, or all three?
• If sanctioned funds landed in a customer’s wallet tomorrow, who would know, how quickly, and who would decide what to freeze?
• What hop distance and risk score do we treat as meaningful, who approved that choice, and have we tested it against real laundering chains?
• Do our partner contracts let us see their alerts and decisions, or only their assurances?
• What does this capability cost us each year, and does the business line justify it?

With hindsight

The legal ground did shift, slowly. Lower-court rulings in 2023 went Treasury’s way. Then, in November 2024, the Fifth Circuit held in Van Loon v. Department of the Treasury that the immutable smart contracts at issue were not “property” that OFAC could block under its statute. The court said, in effect, that code nobody owns or controls is not what the sanctions law was written to freeze. It declined to defer to Treasury’s reading of the term, relying on the Supreme Court’s Loper Bright decision from earlier that year.

A parallel case in Florida, brought by Coin Center, had produced a district court ruling for Treasury in October 2023. In March 2025, Treasury removed Tornado Cash from the sanctions list, and the remaining litigation wound down. Institutions that had blocked balances then faced the exit problem the original designation had created. They had to confirm the change, check for any other sanctions connection, and release funds consistently.

The people behind the software did not escape scrutiny. In May 2024, a Dutch court convicted Alexey Pertsev of money laundering and sentenced him to 64 months; he appealed. In August 2025, a federal jury in Manhattan convicted co-founder Roman Storm of conspiring to operate an unlicensed money transmitting business. The jury could not reach verdicts on the money laundering and sanctions charges. Another co-founder, Roman Semenov, was charged but not tried. The legal system drew a line between the code and the humans who built and ran services around it.

The policy pressure on mixers continued by other routes. In October 2023, FinCEN proposed using its Section 311 authority to treat international crypto mixing as a class of transactions of primary money laundering concern, with new recordkeeping and reporting duties. It was the first time FinCEN had proposed using that authority against a class of transactions rather than a country or institution. OFAC designated another mixer, Sinbad, in November 2023.

North Korean theft did not slow down. In February 2025, the FBI attributed the theft of roughly 1.5 billion dollars in virtual assets from the exchange Bybit to North Korea. OFAC kept listing crypto addresses for designated parties, and blockchain analytics became a standard part of sanctions programs at crypto firms and many banks that serve them. What looked in August 2022 like an exotic one-off turned out to be the start of a permanent feature of financial crime compliance.

The banking backdrop also changed. In March and April 2025, the OCC, the FDIC and the Federal Reserve withdrew the prior-notice requirements and joint statements that had made banks cautious about crypto. In July 2025, the GENIUS Act created a federal framework for payment stablecoins. It requires issuers to be able to freeze and block transactions in response to lawful orders, putting into law the kind of switch Circle had used in 2022. More banks and credit unions now have a route into the business, and with it the on-chain screening questions of 2022.

Three points have lasted. First, a control can be right to run even while its legal basis is being contested; the obligation stood until a court said otherwise, and institutions had to comply in the meantime. Second, documentation is what makes change cheap. Institutions that documented why each control existed found the 2025 delisting a manageable change, a matter of updating configurations and releasing balances in an orderly way. Those that had not were left reconstructing decisions from memory.

What did not hold was the idea that sanctions could reach autonomous code directly. After the Fifth Circuit ruling, at least for immutable contracts, the pressure point moved to people, interfaces and regulated intermediaries. Those were the parts of the system that had carried the real compliance load from the start.

Third, the hardest compliance problems increasingly arise where software, not a person, is the counterparty. That will not be the last time a regulator, a court and a compliance team each try to fit an automated system into rules built for people. The institutions that handled Tornado Cash well were the ones that wrote down their reasoning, set their thresholds deliberately and planned for the rule to change.