TD Bank left more than $100 billion of peer-to-peer payments outside the scenarios
FinCEN’s 10 October 2024 order is a $1.3 billion penalty on TD Bank, N.A. and TD Bank USA, N.A., plus a four-year monitor. Non-Zelle apps were coded as out of scope. Zelle itself waited three years for a code of its own.
Which TD, and which penalty
On 10 October 2024 FinCEN assessed a $1.3 billion civil money penalty against TD Bank, N.A. and TD Bank USA, N.A. The case number is 2024-02. FinCEN called it the largest penalty it has imposed on a depository institution. The order also puts a four-year independent monitorship over the remediation. The two banks are US institutions. They are not, on the face of the order, the whole TD group in Toronto. A questionnaire that says “TD” and does not say which legal entity has already lost the plot. Assets of the US bank, as the order states them, were about $367 billion, and FinCEN describes it as the tenth largest bank in the United States. The same day sat inside a wider resolution. Contemporaneous reporting described further payments to the Justice Department, the Office of the Comptroller of the Currency and the Federal Reserve, and an asset cap, in a package reported at about $3 billion. Faircroft does not add those cheques into a new total. FinCEN’s figure is $1.3 billion. The other figures belong to the agencies that imposed them, cited as reported, and they should not be collapsed into the FinCEN order as if it were one document. What this page uses is the consent order: what the bank admits, what it does not, and which payment rails the monitor could not see. Read the admission carefully, because it is narrower than a headline. The bank admits the facts admitted in the plea agreements of the same day between TD Bank, N.A., TD Bank US Holding Company and the Justice Department. It neither admits nor denies the rest of the narrative in the FinCEN order. It consents to the penalty and to the undertakings. On “willfulness,” it admits the word only in the civil Bank Secrecy Act sense. The order explains that civil willfulness can be reckless disregard or willful blindness, and does not require proof that the bank knew it was breaking the Act or acted with a bad purpose. That is the legal standard FinCEN is using. It is not a quote you should shorten to “the bank confessed.” Quote the order. This was not a first meeting. The order notes a FinCEN action against TD Bank, N.A. in September 2013. A second, much larger order a decade later is a recidivism fact. It does not, by itself, describe the 2024 failures. Those are in the sections on resources, scenarios, and the payment types nobody coded.
The apps that were marked out of scope
Peer-to-peer payments are the part of this order that a payments team can use without reading the whole BSA manual. FinCEN says the bank’s monitoring of peer-to-peer transactions, including activity indicative of human trafficking, was insufficient, and that suspicious activity was not identified and reported in time. The order then splits the rails. Zelle, which the bank launched for personal accounts in April 2017, is not the same pipe as Venmo or PayPal. Because the bank was not monitoring domestic ACH in the way the order describes, peer-to-peer payments made through a channel other than Zelle were not monitored as what they were. Non-Zelle activity — the order lists Apple Cash, Block, Cash App, Facebook Pay, Google Pay, Google Wallet, PayPal and Venmo — was posted as an ordinary domestic debit-card charge or a direct deposit. FinCEN says the bank knowingly failed to monitor appropriately more than $100 billion of those transactions over the period. The paper trail inside the bank is awkward because it is so ordinary. In January 2019 an internal assessment of a terrorist-financing scenario recommended adding online money-transfer systems, including PayPal, Venmo and Zelle, and said that leaving them out reduced the chance of seeing both terrorist financing and other illicit finance. In November 2019, when transaction codes were being updated, those external vendors were marked “out of scope.” The bank never built the codes that would have let a scenario see them. It built codes for Zelle and stopped. In 2021, given another chance to add codes, senior AML staff concluded the existing mechanism was enough. In March 2024, answering a newly hired BSA officer, an AML technology head acknowledged that non-Zelle peer-to-peer payments still had no transaction code that would allow systemic monitoring. Five years after the bank’s own scenario had asked for the coverage, the code did not exist. Zelle was only partly better. From 2017 to 2023 the order counts about 300 million incoming and outgoing Zelle transactions, worth more than $75 billion, roughly as large as all the other peer-to-peer activity combined. Before launch, AML staff had flagged money-laundering risks, including high-risk jurisdictions, and had said the scenarios written for the smaller predecessor product needed another look. After launch they asked for scenarios written for Zelle. The bank did not write them. It dropped Zelle into scenarios built for something else, and for the first years it watched Zelle the way it watched a debit card. Unique codes that could even tell Zelle from other debit-card transactions arrived in April 2020, about three years after launch. Those codes let existing high-velocity scenarios see the flow. They were not scenarios designed for the product. A code is not a typology. It is the precondition for a typology. TD spent three years without the precondition on its own branded rail, and the whole period without it on everyone else’s.
Cheques, a backlog, and a programme that knew it was small
The order’s payment findings do not stop at the phone. In a separate example, a purported real-estate investment scheme from 2018 to 2023 produced more than 3,000 transactions and more than $300 million. Investors were defrauded, on FinCEN’s account. The bank began filing suspicious-activity reports only after a law-enforcement inquiry in 2021, and that reporting covered about 1 percent of the activity. In 2024, more than six years after the activity started, it filed on about 98 percent. Nearly half of the transactions were cheques and cheque deposits, over $40 million, which the bank was not properly monitoring because it did not monitor cheques. A payments desk that thinks of risk as an app will miss the older rail sitting in the same order. Cheques were unscored. Apps were miscoded. The programme’s hole was “we never gave this a code,” not “a clever typology got through.” Underneath both holes is resourcing, which FinCEN describes as known. The bank’s AML programme was not designed or staffed for the risks it actually ran. Alert backlogs persisted, so law enforcement did not get reports in time. Customers sat in an exit backlog. Spending on AML did not rise with the problems the bank had already named, including those backlogs. Training did not equip staff to recognise the activity in front of them, including Ponzi-type schemes. The order’s word for this, in the civil sense above, is willful. A bank that records a recommendation in 2019 and marks the product out of scope the same year has a document. The document is what makes “we did not know peer-to-peer existed” impossible to say.
What a payments team should take from it
Ask, for any sponsor or respondent that offers Zelle, Venmo-type funding, or cheque deposit into the same account: which transaction codes exist, who decided a rail was out of scope, and when that decision was last reopened. Ask for the scenario name that mentions the rail, not the vendor’s marketing name for the app. Ask what share of suspicious-activity reports in a known case were filed after a law-enforcement inquiry rather than before it. The 1 percent figure in the real-estate example is the one to put next to a claim that “we file SARs.” Filing after the government has already asked is a different control from filing when the scenario fires. This page will not say every TD customer who used Venmo was moving illicit funds. More than $100 billion unmonitored is a coverage failure. It is not a finding that the stock of payments was criminal. It will not say the 2013 order and the 2024 order are the same facts. It will say a bank already under a FinCEN order had, a decade later, a peer-to-peer book and a cheque book that the scenarios did not cover. It will not merge the Justice Department plea into this civil order. Read the plea for what was admitted there. Read this order for the rails. Keep the consent order, the code decision of November 2019, and the April 2020 date on which Zelle became distinguishable from a debit card. If your own payment hub posts a third-party app as a generic card purchase, you are running the control FinCEN just priced at $1.3 billion. The fix is a code, a scenario, and a person who is not allowed to mark the rail out of scope without writing down why.