Whether it’s fighting financial fraud or human trafficking, AI is enabling fintech to make a positive impact in crime prevention. 

Where it could be argued that many industries are still wading through onerous self-assessment to find where genAI fits – because, after all, it has to fit somewhere – finance is one area where the technology has been put to work in a way that seems intuitive. 

Fintech technology enables real-time monitoring of financial and investment risks, monetary changes, as well as detecting fraud. Its harnessing of big data analytics is already transforming the financial industry.

The writing’s always been on the wall for these kinds of use-cases of AI in fintech. Where things become a bit murkier is the dichotomy of the technology’s ability to empower fintech to net social benefits (fighting human trafficking and financial fraud) to the other end of the spectrum, where it has lowered the bar for entry, exponentially in some areas, when it comes to participation in crime. 

AI has revolutionised fraud – as the burgeoning technology becomes more easily accessible, fraudsters are better able to add sophistication to their illicit ploys, and unsuspecting consumers are less able to detect social engineering campaigns. AI can write malware, phishing emails, and can even evade detection by other AI. 

But when it comes to fraud detection and prevention, Jamie Hutton, CTO of Quantexa, the AI fintech firm, sticking to the basics is more important than integrating new AI developments.

Back to Basics: A Foundation in Data 

“Of course things like ChatGPT and large language models and other such things have really brought [AI] to the fore, but it doesn’t change the fundamentals,” Hutton said. 

“And the fundamental challenge is around having the data ready for AI.” 

Hutton truly takes it back to basics: “It’s that old adage of garbage in, garbage out. If you don’t have a good data foundation when building your analytics model or your decisioning model on top of, then the machine’s not going to do a very good job because it just doesn’t understand your business or what you’re trying to solve.”

Quantexa has been in utilising AI since its founding back in 2016, and has tracked how the technology and machine learning capabilities have transformed since then, especially in the eyes of the public. 

“Whilst AI has changed, the need for a solid data foundation has absolutely not.”

Applying AI to fraud and crime prevention tactics can allow more data points to be harvested and analysed faster, enabling real-time monitoring and alerts. But with data, quantity does not also equate to greater quality. 

According to Hutton, traditional fraud and money laundering alert systems run at 99% false positives rates. 

This is largely because these, as well as AI systems, only search for anomalies in transactions, without taking into account life circumstances. 

What causes these systems to be so highly inaccurate is a lack of contextual data – banks and financial institutions often know some details about their customers, but do not use other data to place their customers within a relationship network – be it their relationship with other people, bank accounts, or businesses. 

“Rather than just looking at customer behaviour as a very simple thing, we try to understand who the customer truly is,” Hutton said. 

“That’s not so easy,” Hutton admitted. “And of course, if you’re going to solve the fraud of financial crime, knowing where the money flows, is really, really important.”

By harnessing a more varied array of external data, the analytical framework has a better idea of how the real world works – something as simple as connecting with data from Company House to establish that a company is reputable, or connecting a business account to a personal account, can decrease the likelihood of a false positive. 

It also allows a better understanding of the customer’s ‘normal’ behaviour, allowing for more in-depth assessments of potential fraud. 

“The context allows the analytics to understand the real world better. And if the analytics understands the real world, it has a much better chance of being able to detect and prevent fraud,” Hutton said. 

“What our system is trying to do is mirror your best investigator,” Hutton explained. 

By harnessing data to create contextual networks, what Hutton and the team at Quantexa calls entity resolution and network context, the analytics can be inputted into large language models that can then relay risk assessments in digestible language. 

This is where large language models come into play, which are of course, taking the world by storm. 

New Isn’t Always Better

“But where they struggle is with your own internal data,” Hutton conceded. “You need to combine a large language model with your internal data.” 

By combining a robust and varied data network with the processing of large language models, fraud detecting systems can better mimic the processes, and results, of the best investigators, at great speed and efficiency. 

However, as AI comes more into the general foray, so do concerns around its ethical and secure use. 

While AI can be used for fraud detection, how does it mitigate against hallucination, aka when AI generates something with no basis in data or fact, bias, and unexplainable outputs? 

The model Quantexa relies on is transparent and explainable, unlike more recent generative AI algorithms in the foray. 

The philosophy is not so much about finding a new way of doing machine learning, but about inputting better data into a machine learning model. 

The transparency and explainability of the model allows Quantexa to harness data under compliance for fraud and crime detection, mitigate bias, and figure out ways to consistently improve their data input and their algorithm. 

Recommended reading

• FinTech Scotland and Scottish Friendly Announce Strategic Partnership
• New Payments Innovations Report Highlights Sector’s Failures/Successes
• BR-DGE and Volt Partner to Help Merchants Prevent Payment Fraud

The data-driven process has allowed Quentexa to quell some of the more recent fears surrounding AI, and also has enabled it to utilise its model for other applications outside of financial fraud.  

Crime-fighting FinTech

Partnering with the Anti-Human Trafficking Intelligence Initiative (ATII), Quatenxa is applying its Decision Intelligence to the prevention and detection of human trafficking in the US.

To combat this heinous and pervasive crime, the Initiative will deploy Quantexa’s decision making technology as well as its AI-driven crime prevention and detection capabilities through training and joint industry engagement.

The technology essentially works in the same way, but helps not just to detect financial fraud, but illegal financial activity related to human trafficking. 

Clark Frogley, North American Head of Financial Crime Solutions at Quantexa. “Human trafficking is often enabled by illegal financial activity and Quantexa’s deep experience helping organizations to monitor, detect, and investigate financial crimes makes us the ideal partner to support ATII’s mission. This partnership will aid organizations to reinforce their ESG and anti-money laundering (AML) efforts which are critical to stopping human trafficking.”   

In the UK, Quantexa has been helping the Cabinet Office detect fraud, using their AI platform to detect fraudulent businesses taking advantage of the government’s COVID-19 business loan scheme.

For this, Hutton explains the importance of going the extra step to identify fraudulent businesses, harnessing public data and making logical decisions, just as an investigator might. 

“It’s not about just knowing who the people are. You’ve got to look at the corporate structures,” Hutton.

By understanding the corporate layout and history of an organisation, Quantexa is helping the cabinet office better detect fraudulent business owners and fake businesses taking advantage of COVID-19 support.

Recently, Quantexa has expanded their partnership with the Cabinet Office, offering its support to investigate organised crime and sanctions evasions. 

FinTech’s AI Future

Quantexa has been in the AI game since its founding in 2016, and has seen how AI advancements have shifted security trends and challenges in financial services.

“It’s so funny how it shifts over time,” Hutton said, “ I’ve seen huge waves of different types of fraud coming up and down over the time. Obviously one of the biggest ones happening at the moment is things like the scams.

“There’s certainly a challenge around people leveraging AI to convince people to part with their money,” Hutton added. 

AI is allowing people to enhance their scamming techniques, and increases the sophistication of phishing emails. While AI is used defensively, it is increasingly being employed by threat actors.

“You’ve got sort of this adversarial relationship between the detection systems and the criminals, and both are leveraging. AI is going to be a bit of an arms race in many cases,” Hutton predicted.