Artificial Intelligence vs Money Launderers: Who would win?

Nuclear fusion, companionship for the elderly, completing ancient writings, 40,000 new chemical weapons and warfare. These are just a few recent headlines describing the application of Artificial Intelligence (AI) and the sub topic of machine learning in its most nascent of days. With such varied and seemingly fast application, what does it mean for us in the fight against money laundering?

First up, it’s best we agree on what AI is. The encyclopedia Britannica (yes that’s still a thing – online now though) defines it best, “Artificial intelligence (AI), is the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings. The term is frequently applied to the project of developing systems endowed with the intellectual processes characteristic of humans, such as the ability to reason, discover meaning, generalize, or learn from past experience.” 

The relevant phrase for us in the AML space is AI or a machine’s “ability to reason, discover meaning and learn from past experience.” Why’s that you ask? Because although AI is still in its infancy, it has very real implications for tasks that are currently burdensome for humans and/or impossible due to the overwhelming volume of AML transactions and tasks.

In this article we take a look at its current use in AML, relevant application throughout the AML process, possible negative impacts and make some bold predictions about its use in the future. 

AI for AML? 

The Great Reimagination is upon us.

In 2021 the business world was stunned by the sheer volume of resignations, approximately 4% of the working population in the US alone. After 18 months of a pandemic, people were reevaluating their life choices, and jobs that weren’t aligned with life goals were shunned. This led to “The Great Resignation” which led to talent shortages and the resulting talent war. But this has also led to “The Great Reimagination.” 

What started as Industry 4.0 in manufacturing and the Internet of Things (IoT) in the consumer world has spread at an accelerated pace to other business sectors. Business leaders are looking across the globe and to other industries for alternatives that can both alleviate the talent shortage but also give them a competitive edge. They are, in essence, reimagining their businesses. They’re reconsidering what processes could be, what technology does best and where humans add the most value. For us in the AML industry, the opportunities are rich.

Let the bots do the heavy lifting.

Reading and comprehending

The most obvious place to start is data analysis. By training the AI on what information is needed, where to find it and what to ‘read’ using machine vision and supporting comprehension, the manual burden of information collection and verification could be significantly reduced. AI could be used to:

• Find trust deeds
• ‘Read’ deeds to identify beneficial owners to be verified
• ‘Read’ company extracts for ownership structures and owners of the entity
• Define company structures
• ‘Read’ source of wealth documents
• Biometric verification

Volume and complexity

Cyber crime is getting more complex, more brazen and more distributed. Recently we wrote about the impact cyber crime is having on AML, and how easy it is to do,

“​​What used to be the domain of the uber-smart is suddenly available to anyone willing to partner with, and pay, the likes of DarkSide. Small time criminals and ‘wannabe’ hackers now have the world, and its money, at their fingertips.”

So although criminals aren’t necessarily getting smarter, their tools are. And thanks to the world’s hyper connectivity, the entire globe is now a target. 

AML leaders are turning to AI to detect highly suspicious transaction patterns, hidden from humans due to their speed and highly distributed nature. 

As Johanna Walsh, a partner at Kingsley Napley LLP explains, “AI can be particularly useful for detecting unusual single transactions which employees find harder to spot by using systems to build up an image of normal customer behaviour and creating an alert when something unusual arises.”

By leaving the complex and voluminous transaction monitoring to AI it can greatly reduce labour costs and allow AML/compliance specialists to focus on higher value tasks.

This, in turn, improves operational efficiency, streamlining the entire process allowing for more clients to be onboarded while improving quality and accuracy. 

Improve quality and accuracy

AI for AML now

AI can improve quality and accuracy across a number of tasks, but is perhaps best defined by what can be done now, versus what can be done in the near future. 

Electronic Identity Verification (EIV)

Right now, many EIV solutions use AI for: 

• Facial verification
  • Face comparison, user impersonation and liveness testing.

• Document tamper detection
  • Image integrity, relevance and consistency
• Technical tamper detection
  • Metadata analysis, digital tampering detection, time of capture analysis and video texture analysis

Transaction monitoring

As noted earlier, AI is also currently being used to improve quality and accuracy in transaction monitoring by applying a consistent, standardised and well honed approach to compare ‘normal’ transitions to suspicious ones.  

PEPs and sanction lists

The next space that AI is currently being leveraged in, is of significant importance in the current economic environment. AI is helping filter out false positives for politically exposed people (PEP) and sanctioned people while reducing any resulting operational costs.

Again, by utilising multiple sources of verification data, the AI can more consistently, accurately and quickly compare an individual to known PEP and sanction lists – leaving human experts to deal with exceptions and again focus on more value generating tasks.

Reporting

The final use of AI for AML is in reporting. By flagging, triaging and auto generating tasks based on results found in monitoring, verifications and sanctions, compliance teams get better answers more quickly while also meeting regulatory requirements. 

AI for AML later

Reading and comprehension
By far the biggest burden that could be reduced by AI is reading and comprehension. As noted above, the application could offer immense benefit if only the AI-based decision networks and machine reading comprehension existed to make judgment calls on the deeds and other documents they ‘read’.  

Essentially, machine reading comprehension relies on natural language processing and deep learning being established, before introducing the task, models, and applications of it.

Alas, the world’s biggest tech companies, including Microsoft, are in the early days of developing this ability and its commercial application is a way off. 

Concerns, impacts and limitations

AI is in its infancy and as Professor Stuart Russell, founder of the Center for Human-Compatible Artificial Intelligence, at the University of California, Berkeley explains, “I think a little bit of fear is appropriate, not fear when you get up tomorrow morning and think my laptop is going to murder me or something, but thinking about the future – I would say the same kind of fear we have about the climate or, rather, we should have about the climate.”

With that healthy level of fear, we look to the possible problems we should consider in the use of AI in the AML industry.

Bias 

AI is only as good as the datasets it’s trained on. There are stories abound on inbuilt racial bias, sexism and even against those with non-European names. Think of that, how could an AI system treat individuals who didn’t ‘fit’ the training database? 

Security and compliance

As noted at the beginning of this article, criminal tools are getting smarter and cross-border money laundering is relatively easy. To keep data and reputations safe we will need to look at using AI to fight AI-based cyber crime.

Ethics of decisions 

What happens when a ‘bot makes a decision to not accept a 90 year old beneficial owner because they don’t have a driver’s license or a passport to prove their identity? Humans can speak to other humans to understand that the 90 year old has been a customer for 40 years and their advanced age is an obvious reason to not have those identity documents. But AI can’t. 

Correct use cases

As we’ve outlined earlier, AI is only useful in some areas. We still need humans for understanding, intuition and ethical input. Not to mention the ability to engage with other humans to help them through the process, offering support and guidance which AI currently can not.

So you want to try AI

You’ve got through this far and are keen to adopt AI in some form in your AML or compliance programme. That is excellent news. Here’s a brief outline of what to do and how.

• Define your ethical framework that AI decisions are made on
• Identify biases and define mitigation for the AI (see section above)
• Consider how best it integrates with other IT systems
• Choose a non biased, broad data set to train the AI on
• Incorporate geographic-specific regulations and requirements
• Select the right use case for the AI to be applied to
• Run a pilot programme
• Run for iterative improvements / model updates

What’s next

Throughout this article we have noted where AI is currently being used and where it is most likely to be applied in the AML industry. But let’s recap and add a few more thoughts.

AI arms race

We are at the beginning of an AI arms race where, in the parlance, bad actors (cyber criminals) will keep exposing attack vectors (system vulnerabilities) meaning businesses big and small will have to keep up by fighting AI with AI.  

Responsible (ethical) AI 

Just as the public attitude to the climate crisis has evolved, so too will the expectations of AI. Businesses and governments will be expected to ‘do the right thing’ with AI, ensuring the decisions they make are both ethical and responsible for not just now but also in the future. 

Cross border AML collaboration

Although globalisation seems to have recently hit a speed bump, the highly interconnected nature of business, money and economies will still enable global money laundering. With that in mind, we expect broader cross-border AML collaboration to occur, supported by global data sets, utilised by AI for AML crime fighting

AI as standard to protect against identity fraud

This is already happening and will continue to do so. The key is having an exceptional cyber security stance to ensure individuals’ identity data can not be stolen for nefarious uses.

Hybrid approach

The final prediction we make is that for the foreseeable future, AML fighters will use a hybrid AI / human approach to allow for maximum accuracy, improved operational efficiencies and the best customer experience.