GenAI fraud forcing banks to shift from identity to intent

Issue 6 2025 AI & Data Analytics, Information Security, Financial (Industry)

Despite its history of resilience and globally recognised innovation, the eye-watering pace of cyberthreat evolution, characterised by rapid mutation and increasing sophistication, is forcing South Africa’s banking sector to rethink its approach to security.

In its 2024 Annual Crime Statistics report, the South African Banking Risk Information Centre (SABRIC) notes that digital banking fraud saw a significant escalation, with an 86% increase in incidents and a 74% rise in associated losses, totalling R1,89 billion. Digital banking fraud accounts for almost 70% of recorded financial losses in the banking industry. SABRIC has laid a good deal of the blame on emerging technologies, particularly Generative Artificial Intelligence (GenAI), being leveraged by criminals to craft more sophisticated schemes.

South African (SA) banks have not sat idly by; they have invested heavily in systems, partnerships, and awareness campaigns to combat crime. They, too, have turned to AI technologies to fight fraud, saving billions in fraud losses. However, the pace of fraud innovation has made chasing down the criminals a moving target for local banking leaders.

“Gone are the days when a single technology or a static set of controls can keep fraudsters at bay. The threat landscape today is dynamic, with attackers constantly probing for new vulnerabilities and exploiting gaps in traditional defences,” says Pieter de Swardt, SVP Commercial: SA & Sales Operations at Entersekt. “The complexity and velocity of modern fraud schemes, from deepfakes to fraud and scams involving social engineering, demand more than just investment in new tools. While banks are investing heavily in protecting their clients, the real test lies in their ability to adapt and expand their security net.”

A nuanced approach is required

A deliberate move toward multi-layered, adaptive security is required to mitigate these evolving threats. Rather than relying on a single line of defence, banks are increasingly turning to behavioural analytics and risk profiling to analyse intent in addition to identity.

“By constructing dynamic profiles of customer behaviour, including transaction timing, geolocation, device fingerprinting, and channel usage, banks can detect deviations that may signal fraud, even when individual transactions appear legitimate,” says De Swardt.

Banks that focus on whether the client ‘should’ be doing the transaction by analysing context, device usage, and transaction velocity to distinguish genuine behaviour from suspicious intent are reaping the rewards. According to KPMG’s Global Banking Scam Survey 2025, 60% of banks that monitor customers to understand if they are communicating with a third party while using online or mobile banking rated this as an effective fraud prevention measure.

De Swardt explains that shifting to intent-based fraud-fighting approaches also helps fight chargeback fraud, which is a growing headache for banks.

"We had an example of a bank customer reporting fraud after performing multiple transactions in succession. The first transaction was deemed legitimate, but the customer reported the next two as fraudulent. The subsequent transactions were then found to be legitimate. We were able to assist the bank to confirm that all the transactions were coming from exactly the same browser. With the forensics in place, the bank was able to get the client to confess to first-party fraud. The data is impartial and sometimes there is no third party involved,” De Swardt shares.

Cross-channel is key

Equally critical is the need for cross-channel visibility. Fraudsters increasingly exploit gaps between siloed systems, initiating transactions in one channel and authenticating in another.

De Swardt says security architectures must correlate data across originating and authentication channels with online banking, card payments, mobile apps, and other touchpoints. This holistic view enables the detection of anomalous patterns (such as a transaction initiated from an unfamiliar device and authenticated from a geographically distant location) that would be invisible in a single-channel context.

De Swardt says this nuanced approach, using as many data points as possible, allows banks to intervene only when something is suspicious, maintaining a smooth client experience, while still being vigilant against fraud.

“When risk signals indicate that a transaction matches a customer’s usual behaviour, it can be processed quickly and without extra steps. Only when something appears out of the ordinary does the system need to introduce additional authentication measures or alerts, minimising disruption for clients,” he explains.

Security is the one time that competition should not matter

Consortium intelligence is another pillar of effective cyber defence. By participating in industry-wide data sharing initiatives, banks gain access to a broader spectrum of threat intelligence, including indicators of compromise and emerging attack vectors.

“One of the defining features of SA banking is its collaborative spirit. Local banks have established open lines of communication, sharing intelligence and best practices to collectively combat fraud. One of the positive aspects of the battle against fraud is that it is not seen as a competitive advantage. It is a necessity in a region where attack vectors morph rapidly and fraudsters are quick to adapt,” De Swardt shares.

Summing up, De Swardt says the velocity of fraud evolution means that yesterday’s defences can quickly become obsolete. In this context, the technical challenge is not merely to add more layers, but to architect a security ecosystem that is adaptive, intelligence-driven, and capable of real-time response.

“By leveraging behavioural analytics, cross-channel visibility, and industry collaboration, banks can stay ahead of evolving threats while preserving the trust and satisfaction of their customers. The battle against cybercrime is only beginning, but with a new approach, as well as the right partners, SA banks are well positioned to meet the challenge,” he says.

For more information, go to www.entersekt.com




Share this article:
Share via emailShare via LinkedInPrint this page



Further reading:

Who are you?
Access Control & Identity Management Information Security
Who are you? This question may seem strange, but it can only be answered accurately by implementing an Identity and Access Management (IAM) system, a crucial component of any company’s security strategy.

Read more...
Check Point launches African Perspectives on Cybersecurity report
News & Events Information Security
Check Point Software Technologies released its African Perspectives on Cybersecurity Report 2025, revealing a sharp rise in attacks across the continent and a major shift in attacker tactics driven by artificial intelligence

Read more...
What is your ‘real’ security posture?
BlueVision Editor's Choice Information Security Infrastructure AI & Data Analytics
Many businesses operate under the illusion that their security controls, policies, and incident response plans will hold firm when tested by cybercriminals, but does this mean you are really safe?

Read more...
What is your ‘real’ security posture? (Part 2)
BlueVision Editor's Choice Information Security Infrastructure
In the second part of this series of articles from BlueVision, we explore the human element: social engineering and insider threats and how red teaming can expose and remedy them.

Read more...
IQ and AI
Leaderware Editor's Choice Surveillance AI & Data Analytics
Following his presentation at the Estate Security Conference in October, Craig Donald delves into the challenge of balancing human operator ‘IQ’ and AI system detection within CCTV control rooms.

Read more...
New agent gateway to mitigate shadow MCP risk
AI & Data Analytics
Agent Gateway, a new capability in the Tray AI Orchestration platform, gives IT power to develop approved MCP tools with policies, permissions, versioning and compliance, then publish them via MCP for secure agent use.

Read more...
AI rewrites financial crime
Security Services & Risk Management Financial (Industry)
Criminals are exploiting South Africa’s high connectivity and still-maturing regulation to scale attacks faster than we can defend them. The speed and sophistication of these scams are outpacing the systems designed to stop them.

Read more...
AI and automation are rewriting the cloud security playbook
Technews Publishing AI & Data Analytics
Old-school security relied on rules-based systems that flagged only what was already known. AI flips the script: it analyses massive volumes of data in real-time, spotting anomalies that humans or static rules would miss.

Read more...
Onsite AI avoids cloud challenges
SMART Security Solutions Technews Publishing Editor's Choice Infrastructure AI & Data Analytics
Most AI programs today depend on constant cloud connections, which can be a liability for companies operating in secure or high-risk environments. That reliance exposes sensitive data to external networks, but also creates a single point of failure if connectivity drops.

Read more...
neaMetrics 2025: Year in review
AI & Data Analytics
With a stronger team, a broader portfolio, and a clear vision for what’s next, neaMetrics is well positioned to continue delivering smarter, more connected security solutions in 2026 and beyond.

Read more...










While every effort has been made to ensure the accuracy of the information contained herein, the publisher and its agents cannot be held responsible for any errors contained, or any loss incurred as a result. Articles published do not necessarily reflect the views of the publishers. The editor reserves the right to alter or cut copy. Articles submitted are deemed to have been cleared for publication. Advertisements and company contact details are published as provided by the advertiser. Technews Publishing (Pty) Ltd cannot be held responsible for the accuracy or veracity of supplied material.




© Technews Publishing (Pty) Ltd. | All Rights Reserved.