Blockchain analysis firm Elliptic, MIT, and IBM have released a new AI model—and the 200-million-transaction dataset it’s trained on—that aims to spot the “shape” of bitcoin money laundering.
Imagine you’re a detective with a magnifying glass, but instead of looking for clues in a dusty old attic, you’re exploring a giant digital world where everyone uses secret nicknames. This world is called the Bitcoin blockchain, and it’s filled with nearly a billion secret messages that tell the story of money moving from one place to another.
Now, some people in this world might not be playing by the rules; they could be trying to hide or move stolen money. To catch them, we use a super-smart helper—artificial intelligence (AI), which is like a robot detective. This robot detective is really good at finding patterns and clues in the giant digital world that most people, or even regular detectives, can’t see.
Recently, a group of smart scientists from a company called Elliptic, along with their friends at MIT and IBM, decided to teach a new robot detective how to spot these rule-breakers. Instead of looking for bad guys one by one, they looked at how money moved in groups of messages, sometimes up to six at a time, from the bad guys to places where they turn the stolen money into cash.
The scientists collected lots of these secret messages and showed them to their robot detective. By studying these, the robot learned to recognize the special patterns that might mean someone is laundering money—this means they are trying to make bad money look good.
What’s really exciting is that the scientists shared their robot detective’s notebook, which has 200 million of these secret messages, with everyone! That’s like giving every detective in the world a copy of the most detailed clue book ever, making it a thousand times better than any clue book before. This big book of clues doesn’t just point out the bad guys; it shows patterns and chains of clues that help trace the bad money back to its source.
A few years ago, in 2019, these scientists did something similar but on a much smaller scale. They trained their robot with fewer messages, but this time, they went all out to make their robot even smarter.
So, think of this whole thing as a giant game of detective, where the robot is getting smarter and the book of clues is getting bigger, all to help catch the rule-breakers in the digital world and keep the game fair for everyone.
n the vast digital landscape of the Bitcoin blockchain, even the smallest breakthrough in detecting financial misconduct can have massive implications. Recently, a team of researchers at Elliptic, MIT, and IBM made such a breakthrough with their latest AI tool. This new AI is trained to sift through millions of transactions to find patterns that suggest money laundering. While finding 14 suspicious accounts out of 52 might not seem impressive at first glance, it’s important to understand the context. Typically, only 0.1 percent of accounts on an exchange are flagged for potential money laundering. This AI tool, however, has honed in on more than one in four accounts, a significant leap from the usual “one in a thousand.”
Mark Weber, a coauthor of the study and a fellow at MIT’s Media Lab, explained the significance of their findings with enthusiasm: “Going from ‘one in a thousand things we look at are going to be illicit’ to 14 out of 52 is a crazy change.” This enhanced detection rate not only narrows down the pool of potential suspects drastically but also allows investigators to focus their efforts more effectively, ensuring that fewer illicit activities slip through the cracks.
The real-world impact of this AI tool is already becoming evident. Elliptic has been using the model in its operations, uncovering significant leads. For instance, the tool helped trace the source of funds for several suspicious transaction chains. This investigation led to the discovery of Bitcoin addresses linked to a notorious Russian dark-web market, a cryptocurrency “mixer” used to obscure the origins of Bitcoin transactions, and even a Panama-based Ponzi scheme.
These discoveries highlight the tool’s ability to dig deeper into the blockchain and unravel complex chains that might otherwise remain hidden. By providing such detailed insights, the AI model not only aids in immediate investigations but also helps in understanding and dismantling sophisticated financial networks designed to evade traditional detection methods.
As this technology continues to evolve and improve, the possibilities for its application in financial monitoring and law enforcement are vast. The capability to rapidly identify and investigate suspicious accounts promises a new era in the fight against cryptocurrency-based crime, making the digital financial world a safer place for legitimate users and a challenging one for perpetrators.