The Fraud Examiner

AI in the Fight Against Fraud

Mason Wilder, CFE                            
Research Specialist, Association of Certified Fraud Examiners                                 

While the mention of artificial intelligence (AI) may prompt many people to immediately visualize sentient robots taking over Earth and subjugating the human race, limited forms of AI have already been contributing to anti-fraud efforts for decades. As AI tools evolve, so too will their roles in fighting fraud, with major implications for many industries in which fraud is prevalent.

As technology shapes the global economy and consumer behavior, it also influences the way fraud is perpetrated, thereby requiring adaptation in the perpetual fight against fraud. One tool that many organizations are looking to use is AI.

A brief history of AI

Before going too far into AI’s current and future fraud-related applications, it might help to explain some concepts crucial to understanding AI and how it works. Many CFEs have surely heard the phrases “machine learning,” “neural networks,” “deep learning” or “natural language processing,” but what do they mean and why do they matter in a fraud context?

In the early 1940s, scientists first conceptualized a model for an artificial neural network based on the biological neural network of a human brain and began applying it to computational machines, such as calculators, in the 1950s. These concepts and applications were primarily focused on machine learning, or computers recognizing patterns and making calculations or predictions based on those patterns, and would lay the groundwork for all future AI. Progress stagnated in the early stages of machine learning, however. Artificial neural networks, made up of many layers of artificial neurons and synapses, required vast amounts of computer processing power at a time when basic computers took up entire rooms.

Decades later, computer processing power had made significant strides, and computer programmers were able to write programs with well-defined rules for computers to apply in analyzing large sets of data. In 1987, Security Pacific National Bank became the first financial institution to apply this basic machine learning technology, or AI, to preventing fraud. They created a system incorporating the industry’s understanding of fraud indicators into rules that would identify possible fraud in debit card transactions and alert the bank.

These early rules and reputation-based AI systems were applied in other industries as developers continued working on pushing computers’ machine learning capabilities to further streamline analysis of ever-growing data sets. The goal for anti-fraud applications of AI would be to create systems that could go beyond flagging potentially fraudulent transactions to learning from the data those systems process.

Sign In

Not a member? Click here to Join Now and access the full page.