AI meets Forensics

AI is changing nearly every industry, and it will have a major impact on our society in ways we still barely understand. Use cases for the application of AI in general, and Generative AI specifically, seem to be endless. There is no doubt that these developments in AI will transform fraud and financial crime (F&FC), both how it is committed and how we try to prevent, detect and respond to this threat. In this blog, the first in a series of AI-powered F&FC blogs, we will explore how AI is changing the game in the F&FC space.

In the F&FC space, AI is not new, with perhaps the most well-known or most effective example being credit card fraud detection. In 1992, the Falcon Fraud Manager already applied a neural network-type detection system to detect fraudulent payments. Since then, AI has emerged as a transformative force, supporting forensic professionals in the prevention and detection of, and response to fraud. However, not all initiatives are as successful as credit card fraud detection. Why is that? Unfortunately, AI has also given fraudsters the tools to commit fraud more easily, but perhaps there are other reasons as well, as to why AI isn’t the panacea.

In this first blog, we will highlight how AI can be used within the domain of fraud and financial crime, as well as how, specifically, the misuse of AI to commit fraud can be combatted. The other blogs in this series will provide deep dives into all of these topics. So stay tuned!

AI equips fraud specialists with a set of robust tools and controls to prevent, detect and respond to fraud. Here is how AI augments the work of forensic professionals:

1. Advanced algorithms and behavioral analytics. Retail theft has been on the rise over the past few years, and also from a corporate security perspective there are rapid developments in how we work, that impact the effectiveness of security measures. This is where Fraud Data Analytics (FDA) combined with Artificial Intelligence (AI) comes into play. By leveraging advanced analytics and AI, organizations can detect and prevent fraudulent activities more effectively. How can organizations start experimenting with such techniques in a responsible way?

2. Purpose-based KYC and AML monitoring. Current-day compliance activities sometimes seem to lack purpose, with a significant part of the efforts being used, through KYC and transaction monitoring activities, to evaluate people who pose a limited risk. Analytics and AI have been used to improve these processes but not yet with sufficient effect. What perspective does GenAI offer to change how financial institutions handle the risk of money laundering and fraud, and what are potential drawbacks and limitations?

3. eDiscovery powered by GenAI. Roughly 360 billion emails are sent and received per day, and on top of that we use communication tools like Teams, WhatsApp and Slack. When conducting an eDiscovery investigation, for example to investigate a fraud, we want to be able to identify the ‘needle in the haystack’. With this increase in communication, the haystack keeps growing. Utilizing GenAI can help in this search, by making it more efficient (reducing manual workload) and more effective (covering more data, without impacting data privacy, and using opened up time for a second review). So, what is the current state of GenAI-powered eDiscovery, and what are the risks?

Use cases for fraudsters to use GenAI have already made headlines in the media. Websites like FraudGPT and OnlyFake make it easy to commit fraud, in the same way that in the mid-90s script kiddies were able to hack computers, simply by downloading specialized tools. Additionally, GenAI also helps the seasoned fraudster, with elaborate and targeted attacks using deepfake videos, synthesized audio and email communication in the style of a CEO. But AI can also help fraud specialists up their game. This development will once again lead to a cat-and-mouse game.

1. Identification of AI-generated content: In today’s world, it is still generally possible to spot scam messages, photoshopped IDs or fake and adjusted loan applications, if you pay enough attention. GenAI-generated content can be so sophisticated and convincing that you will need to use AI to spot AI. What can we do to not lose this cat-and-mouse game?

2. Model Validation of AI systems: Although not deliberately, the use of AI and GenAI by organizations can result in misuse of technology. Regulators have worked on the recent AI Act, and several regulators have also issued guidance to help organizations make responsible use of AI while evaluating and controlling the resulting risks. But how would this look in practice, within the F&FC space? And how does this differ from techniques already in place, such as transaction monitoring, which already assess large amounts of data?

Additionally, while fraud specialists may not share the same malicious intent as criminals or fraudsters, AI can still be unintentionally misused when applied to fraud prevention and detection. This is why the responsible implementation, usage, and ongoing monitoring of AI are critical. In our final blog of this series, we will explore the compliance considerations and ethical implications of using AI in fraud management.

These AI developments within the F&FC space are not just a glimpse into the future but a pivotal paradigm shift in combating fraud. Fraud specialists need to harness the (predictive) power of AI, if they do not want to fall behind; embracing AI-driven solutions is not merely an option but a necessity to safeguard businesses, economies, and stakeholders against the ever-evolving threats of fraud. With the coming blogs in this series, we want to help you stay ahead in the relentless battle against fraud.