The pandemic has changed how criminals hide their cash—and AI tools are trying to sniff it out
When economies across the world shut down earlier this year, it wasn’t only business owners and consumers who had to adapt. Criminals suddenly had a problem on their hands. How to move their money?
Profits from organized crime are typically passed through legitimate businesses, often exchanging hands several times and crossing borders, until there is no clear trail back to its source—a process known as money laundering.
But with many businesses closed, or seeing smaller revenue streams than usual, hiding money in plain sight by mimicking everyday financial activity became harder. “The money is still coming in but there’s nowhere to put it,” says Isabella Chase, who works on financial crime at RUSI, a UK-based defense and security think tank.
The pandemic has forced criminal gangs to come up with new ways to move money around. In turn, this has upped the stakes for anti-money laundering (AML) teams tasked with detecting suspicious financial transactions and following them back to their source.
Key to their strategies are new AI tools. While some larger, older financial institutions have been slower to adapt their rule-based legacy systems, smaller, newer firms are using machine learning to look out for anomalous activity, whatever it might be.
It is hard to assess the exact scale of the problem. But according to the United Nations Office on Drugs and Crime, between 2% and 5% of global GDP—between $800 billion and $2 trillion at current figures—is laundered every year. Most goes undetected. Estimates suggest that only around 1% of profits earned by criminals is seized.
And that was before covid-19 hit. Fraud is up, with fears around covid-19 creating a lucrative market for counterfeit protective gear or medication. More people spending time online also creates a bigger pool for phishing attacks and other scams. And, of course, drugs are still being bought and sold.
Lockdown made it harder to hide the proceeds—at least to begin with. The problem for criminals is that many of the best businesses for laundering money were also those hit hardest by the pandemic. Small shops, restaurants, bars, and clubs are favored because they are cash-heavy, which makes it easier to mix up ill-gotten gains with legal income.
With bank branches closed, it has been harder to make large cash deposits. Wire transfer services like Western Union—which usually allow anyone to walk in off the street and send money overseas—shut their premises, too.
But criminals are nothing if not opportunistic. As the normal channels for money laundering closed, new ones opened up. Vast sums of money have started flowing into small businesses again thanks to government bailouts. This creates a flurry of financial activity that provides cover for money laundering.
Breaking the rules
The upshot is that there are more demands being placed on AML tech. Older systems rely on hand-crafted rules, such as that transactions over a certain amount should raise an alert. But these rules lead to many false flags and real criminal transactions get lost in the noise. More recently, machine-learning based approaches try to identify patterns of normal activity and raise flags only when outliers are detected. These are then assessed by humans, who reject or approve the alert.
This feedback can be used to tweak the AI model so that it adjusts itself over time. Some firms, including Featurespace, a firm based in the US and UK that uses machine learning to detect suspicious financial activity, and Napier, another firm that builds machine learning tools for AML, are developing hybrid approaches in which correct alerts generated by an AI can be turned into new rules that shape the overall model.
The rapid shifts in behavior in recent months have made the advantages of more adaptable systems clear. Financial regulators around the world have released new guidance on what sort of activity AML teams should look out for but for many it was too late, says Araliya Sammé, head of financial crime at Featurespace. “When something like covid happens, where everybody's payment patterns change suddenly, you don’t have time to put new rules in place.”
You need tech that can catch it as it is happening, she says: “Otherwise by the time you’ve detected something and alerted the people who need to know, the money is gone.”
For Dave Burns, chief revenue officer for Napier, covid-19 caused long-simmering problems to boil over. “This pandemic was the tipping point in many ways,” he says. “It's a bit of a wake-up call that we really need to think differently.” And, he adds, “some of the larger players in the industry have been caught flat-footed.”
But that doesn’t simply mean adopting the latest tech. “You can’t just do AI for AI’s sake because that will spew out garbage,” says Burns. What’s needed, he says, is a bespoke approach for each bank or payment provider.
AML technology still has a long way to go. The pandemic has revealed cracks in existing systems that have people worried, says Burns. And that means that things could change faster than they were going to. “We’re seeing a greater degree of urgency,” he says. “What is traditionally very long, bureaucratic decision-making is being accelerated dramatically.”
Geoffrey Hinton tells us why he’s now scared of the tech he helped build
“I have suddenly switched my views on whether these things are going to be more intelligent than us.”
ChatGPT is going to change education, not destroy it
The narrative around cheating students doesn’t tell the whole story. Meet the teachers who think generative AI could actually make learning better.
Deep learning pioneer Geoffrey Hinton has quit Google
Hinton will be speaking at EmTech Digital on Wednesday.
We are hurtling toward a glitchy, spammy, scammy, AI-powered internet
Large language models are full of security vulnerabilities, yet they’re being embedded into tech products on a vast scale.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.