You may already be familiar with artificial intelligence (AI) and even use it in business applications. What about generative AI (Gen AI)? Gen AI typically is used to create new content (such as text, images, code and video) from raw data and includes such popular systems as Grok, ChatGPT and DALL-E.

Gen AI can also help reduce, uncover and combat fraud. However, there may be ethical and legal implications of using these applications.

Digital decisions

Fraud can occur when a perpetrator finds one or more weaknesses in a company’s defenses and exploits them, creating patterns of activity. Finding these patterns quickly is critical to minimizing losses. Traditional fraud detection engines rely on rules to detect suspicious activity. Once the technology detects a transaction that fits a pattern, the activity is queued for review by a fraud investigator, who then decides whether the activity is indeed fraudulent.

Fraud detection powered by AI uncovers activities that fit previously identified fraud schemes and can evolve and develop new rules based on emerging patterns and trends. So in addition to detecting tried and tested fraud schemes, AI can find suspicious transactions as fraudsters alter their schemes.

Gen AI can automate approval or rejection of potentially fraudulent transactions, thus eliminating the need for manual review by an investigator. It makes “digital decisions” using a combination of rules, a company’s policies and procedures, and the technology’s ability to analyze and learn from real-time data.

Hidden connections

Combatting fraud requires analyzing vast amounts of data quickly and accurately. Fraud detection can become extremely time-consuming and error-prone without sophisticated technology. Fraud investigators often become overwhelmed with “false positives” — transactions that appear fraudulent but, on closer inspection, are legitimate. Human fatigue is also a factor. But Gen AI minimizes false positives and ensures consistent handling of suspicious transactions.

It also can uncover anomalies that might otherwise go undetected. Take, for example, a series of fraudulent transactions perpetrated over an extended period involving accounts connected to the same street address. It would be almost impossible for a human to detect them, especially if they occur over many months or years. Gen AI can detect the commonalities and flag subsequent transactions that follow the same pattern.

The success of any fraud detection solution depends on its ability to detect the latest evolution of fraud. Gen AI helps by 1) anticipating how a fraud scheme might evolve and then building rules to detect its occurrence and 2) using established schemes to produce synthetic data to train and optimize a fraud detection solution. Tasking Gen AI with developing synthetic data ensures your company remains at the cutting edge of fraud prevention.

Ethical challenges

The law governing Gen AI is still evolving, and it may be several years before there’s clear guidance for companies to follow. But as with any form of technology, there’s a potential for unintended outcomes, including implicit bias and privacy violations.

Before deploying Gen AI, understand how the technology uses data. Ask your solution provider to explain how it ensures the protection of your data from unauthorized use, how it complies with current laws and regulations, and how it plans to comply as new ones are enacted.

(This is Blog Post #1525)