Artificial Intelligence in AML/CFT: A Criminological Evidence Map and Multilevel Governance Framework for Organised Financial Crime Control
This article reframes AI in anti-money laundering and counter-terrorist financing as a criminological dual-use challenge rather than a mere efficiency tool, synthesizing evidence to reveal how AI simultaneously enhances detection capabilities and empowers criminal actors, thereby necessitating a comprehensive multilevel governance framework that balances opportunity reduction, legal accountability, and rights protection.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
The Digital Cat-and-Mouse Game: Why AI is a Double-Edged Sword in the Fight Against Crime
Imagine the global financial system as a massive, bustling city where money flows like traffic. In this city, there are two main groups of drivers: the law-abiding citizens who need to move their cash for business and charity, and the "shadow drivers"—criminals and terrorists who are trying to hide their vehicles, change their license plates, or sneak illegal cargo through checkpoints. For decades, the city's security guards (banks and regulators) have used rulebooks to spot these shadow drivers. If a car moves too fast, stops in a suspicious place, or looks like a known criminal vehicle, the guards pull it over. But the shadow drivers are getting smarter, using complex maps and disguises to slip through the cracks.
Enter Artificial Intelligence (AI). Think of AI as a super-powered, high-tech radar system that can spot patterns invisible to the human eye. It can connect dots between thousands of cars instantly, predicting which ones might be carrying contraband. But here's the twist: the shadow drivers are also buying this same super-radar. They are using it to design better disguises, forge fake ID cards, and trick the security guards. This isn't just a story about faster computers; it's a story about a high-stakes game of "cat and mouse" where both sides are upgrading their toys at the same time. The big question isn't just "Can AI catch the bad guys?" but "Can we use AI without accidentally locking up the innocent or letting the pros escape?"
The Paper's Story: A Map of the Digital Battlefield
In this research article, authors Raymond Tang and Bo Wen act like detectives mapping out this high-tech battlefield. They aren't just looking at how fast AI can count money; they are asking a much deeper question: How does AI change the game of crime itself? They treat Anti-Money Laundering (AML) and Countering the Financing of Terrorism (CFT) not just as a math problem, but as a crime-control problem where criminals and regulators are constantly evolving together.
The authors built a "evidence map," which is like a treasure map showing where we have solid gold (real proof) and where we only have rumors (ideas). They looked at thousands of documents, from computer science papers to government laws, to see what actually works and what is just hype.
The Good News: Where the Radar Works
The paper finds that we have some strong evidence for certain AI tools, but they are mostly in the "training wheels" or "simulation" phase, not fully proven in the real world yet.
- The Pattern Hunters (Machine Learning): Imagine a detective who has read every crime novel ever written. AI can learn from past financial records to spot weird patterns that humans miss. The paper says this is great for prioritizing alerts. Instead of a human checking 1,000 suspicious transactions, the AI says, "Hey, check these 10 first!" However, the authors warn that this is mostly based on Level 2 evidence (simulations and controlled tests). It hasn't been fully proven to work perfectly in every real bank yet.
- The Network Connectors (Graph Analytics): Money laundering often looks like a spiderweb. Criminals move money through dozens of shell companies and fake accounts. Graph analytics is like a tool that draws lines between all these dots, revealing the whole web. The paper finds this is very promising for spotting organized crime networks, but again, most of the proof comes from Level 2 (simulations) or Level 3 (banks trying it out in pilots). We don't have enough Level 4 (independently proven, real-world success) stories yet.
- The Crypto Trackers (Blockchain Analytics): Since digital currencies (like Bitcoin) leave a public trail, AI can follow the money across the internet. The paper notes this is useful, but it's tricky. Criminals use tools to hide their tracks, and the AI isn't always perfect at unmasking them.
The Bad News: The Criminals Are Using AI Too
This is the most critical part of the story. The paper argues that AI is a dual-use tool, meaning it helps the good guys and the bad guys equally.
- The Deepfake Disguise: Criminals are using "Generative AI" (the kind that writes stories and makes images) to create fake IDs, deepfake videos for video calls, and realistic fake documents. This makes it incredibly hard for banks to know if the person opening an account is real. The paper suggests that while we think AI can catch these fakes, the criminals are getting better at making them faster than we can catch them.
- The "Synthetic" Lie: Criminals are using AI to create "synthetic identities"—fake people made from a mix of real and fake data. It's like a criminal building a ghost that looks real enough to trick a bank's computer.
- The Script Writers: AI can write scripts to recruit "money mules" (people who move dirty money for a cut) or write fake business reasons for why a huge transfer is happening.
The Tricky Part: When the Radar Goes Wrong
The paper shines a light on a dangerous side effect: De-risking. Because AI systems are sometimes too scared of making mistakes, they might decide it's safer to just ban entire groups of people.
- The Humanitarian Problem: The authors warn that AI might accidentally flag legitimate charities, refugees, or families sending money home (remittances) as "terrorists" just because they are from a certain country or use a certain name. If the AI sees a pattern that looks like a "risk," it might shut down the account. The paper argues this is a huge problem because it hurts innocent people and pushes money into the shadows where it's even harder to track.
- The "Black Box" Issue: Sometimes, the AI says, "This is suspicious," but it can't explain why. If a bank can't explain to a customer why their account was frozen, it breaks the rules of fairness and law.
The Solution: A Multi-Level Game Plan
Since we can't just turn off the AI, the authors propose a Multilevel Governance Framework. Think of this as a set of rules for the whole city, not just the banks.
- For Banks (The Guards): Don't just trust the AI. Keep a human in the loop. If the AI flags something, a real person must check it. Also, banks need to test their AI constantly to make sure it isn't biased against certain groups.
- For Crypto Companies (The New Drivers): They need to follow strict rules about who they let in and how they track money, especially with "stablecoins" (digital money that stays the same value).
- For Regulators (The City Planners): They need to understand the technology. They can't just ask for a report; they need to be able to test the AI themselves to see if it's actually working or just pretending.
- For Everyone: We need to protect human rights. If the AI makes a mistake, there must be a way for a person to appeal and get their account back.
The Bottom Line
The paper concludes that AI is not a magic wand that will solve money laundering overnight. It is a powerful tool that makes the game faster and more complex for everyone. The authors suggest that we are in a co-evolutionary contest: as we build better AI to catch criminals, criminals build better AI to hide.
The most important takeaway is that we must be careful. We can't just let the AI run the show. We need human accountability, fairness checks, and clear rules to make sure that while we are catching the bad guys, we aren't accidentally locking up the good ones or letting the criminals win by using our own tools against us. The paper doesn't claim to have solved the problem, but it provides a clear map of where we stand and what we need to do next to keep the financial city safe.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.