Scams, TD Bank + Losing the Arms Race

What’s up, everyone – Pranjal here.

We’re back with another edition of Generative Finance - the best in finance x AI news.

This week we’re talking about TD Bank’s big fine and the arms race that we’re losing in compliance right now.

Before we dive in, a quick note. I’m heading to Vegas for Money 2020 at the end of October. If you’re also attending, message me.

Would love to meet!

My favorite finds of the week.

• We need to talk about scams (link)
• Innovation may slow amid bank-fintech partnership scrutiny (link)
• A month in the life of a compliance officer (link)
• First & Peoples currently in dispute with bankrupt "fintech" lender (link)
• Revolut doubts Meta’s anti-scam plan (link)

NEWS

TD Bank’s $3 billion compliance catastrophe

TD Bank, one of America's largest banks, just got slapped with a record-breaking $3 billion penalty by FinCEN for chronic violations of anti-money laundering laws. This financial fumble is more than just a number—it's a wake-up call for the entire banking industry.

The backstory: For over a decade, TD Bank let its AML program gather dust, becoming a magnet for illicit actors—including its own employees. The bank's lax oversight allowed trillions of dollars in transactions to slip through unchecked, failing to report thousands of suspicious activities totaling about $1.5 billion.

Now... TD Bank is facing the music with the largest penalty ever imposed on a U.S. depository institution. But the punishment doesn't stop at the price tag:

1. A four-year independent monitorship will oversee TD Bank's required remediation.
2. The bank must conduct a "SAR lookback" to file all those missed Suspicious Activity Reports.
3. For the first time, FinCEN is mandating accountability and data governance reviews.

The crux of the issue? TD Bank's failures weren't just technical glitches—they were systemic breakdowns that facilitated everything from fentanyl trafficking to terrorist financing. In one jaw-dropping example, the bank allowed over $400 million in suspicious transactions for a single individual who later pleaded guilty to money laundering.

THE TAKEAWAY Complacency is criminality. TD Bank's billion-dollar blunder is a stark reminder that robust AML programs aren't just checkboxes—they're safeguards against real-world harm. As regulators sharpen their teeth, banks must realize that the cost of cutting corners on compliance far outweighs the investment in robust systems and vigilant oversight.

Plus, we’re living in an era where peer-to-peer transactions and digital banking are the norm. And, when money moves faster… scams do, too. Financial institutions have to evolve their compliance strategies at the same speed or risk becoming accidental accomplices to the kinds of crimes they're meant to prevent. After all, it is really hard to hire your way out of a hole caused by a lack of investment in tech.

MY TAKE

There’s a new arms race… and we’re losing

The U.S. Treasury Department had an eventful summer. For the last few months, it’s been soliciting comments on AI from financial services companies and regulators. The Consumer Financial Protection Bureau (CFPB) posted their comment a few weeks ago, giving us a peek inside the Bureau’s thought process around AI.

Here are the highlights:

1. The use of AI by financial institutions falls under existing laws and regulations (aka we regulate behaviors and actions, not the tech powering them)
2. AI should be anti-discriminatory and complex models should be regularly tested for disparate treatment (e.g. to maintain fair lending protocols)
3. The regulation of AI should foster competition by creating a level playing field

I want to talk about the last point, which incidentally is super relevant to the news about TD Bank this week. I think we’re focused on leveling the wrong **playing field. (Not to say that I disagree with the CFPB’s note - competition between institutions is of course a good thing!).

But while the Bureau is making sure complex bank models don’t discriminate against your credit score – again, a very worthwhile aim – the dark web is teaching AI how to mimic your grandma’s voice and sweet-talk their way into her retirement account.

This is where the real asymmetry lies: between the good guys (banks and fintechs) and the bad guys (criminals).

Both sides have access to complex AI models, but only one side has to jump through regulatory hoops – and, spoiler alert, it’s not the criminals. Without any legal speed brakes slowing them down, the criminals targeting both individuals and financial institutions are able to move a lot faster and are so much harder to outsmart.

For example, there’s mounting evidence that AI can help them manipulate markets by executing trades at lightning speed. Or can help create fake transactions that fool compliance tools. In one case, deepfake technology was used to scam a UK-based energy firm out of ?220,000. Criminals mimicked the CEO's voice so convincingly that the company unknowingly transferred the funds to a fraudulent account.

Nick Sharp, deputy director at the UK's National Economic Crime Centre (NECC), recently raised a red flag at a London conference. He warned that while AI hasn't yet been widely deployed in fraud, it's “probably the most significant risk facing us, and there’s going to be an innovation arms race.”

What we don't want is for criminals to outpace us, and for regulators and compliance teams to always be stuck playing catch-up.

Evening this playing field, like Nick says, is going to take a lot of innovation. This doesn’t mean loosening regulations and calling AI a free-for-all. It means financial institutions being willing to look at their processes closely, identify where all the easily-exploitable weak links are, and build over them. The weak links might be in error-prone manual work, lack of education around new scams, or in new tech that’s underutilized.

When me and my co-founder, Yutong, were at Brex, we worked in a team that reduced millions of dollars in fraud losses. We had two consistent learnings from that time:

1. Fraud is agile
2. Too many organizations (even the big banks!) use outdated tech, or no tech at all, to combat very real threats.

The reality is, yesterday’s tried-and-tested techniques aren’t going to work with today’s type of crime. We need compliance processes to be as agile and that evolve with with the fraud they’re trying to target.

After all, you can’t bring a knife to a gun fight and expect to win (case in point: TD Bank).

Until next time,

Pranjal

How I can help

We can help speed up your compliance and onboarding process.

We built Accend as your AI-powered platform to help risk and compliance teams get customers onboard quicker. Get started today with us today.