Phishing scams don’t look like they used to. Scammers have done away with typos and obvious fake addresses, replacing them with polished, error-free messages that are getting increasingly difficult to spot thanks to GenAI. As AI continues to lower the barrier to entry for cybercriminals, the sheer scale and sophistication of attacks have left customers concerned about losing money, control of their accounts, and even their identity.
To safeguard the integrity of financial institutions and preserve customer trust, the Payment Systems Regulator (PSR) has introduced new regulations requiring banks to reimburse authorised push payment fraud (APPF) victims up to £85,000 within five days.
Effective from October 7th, new regulations under the PSR mandate a 50/50 liability split between sending and receiving payment services providers for fraud reimbursement. Unlike the optional Contingent Reimbursement Model (CRM), which was previously in place, this mandatory regulation shifts the responsibility squarely onto fraud and cyberfusion teams to stay ahead of evolving threats. With the ease and speed of real time payments, deploying effective fraud detection technology that seamlessly integrates without disrupting the customer experience is now a top priority.
How AI has changed the game for fraudsters
In 2022, there were an estimated 4.7 million phishing attacks in the UK, according to The Independent. Thanks to AI, scammers can now distribute sophisticated attacks across multiple channels - including SMS texts, QR codes, and even voice calls using AI-enhanced voice-altering technology. This advanced automated attack technology is available on the surface web and new recruits are being encouraged to join in through effective social media campaigns highlighting the FOMO (feeling of missing out) factor. As a result, fraud attempts are not only becoming more frequent but also harder to spot.
As fraud tactics evolve, it's clear that traditional approaches to fraud detection are no longer enough. So, what should banks look for when selecting an effective fraud detection solution?
Where do some fraud detection solutions fall short?
The first mistake banks often make is underestimating how fluid and unpredictable financial activity can be. Unsupervised fraud detection models are designed to flag ‘unusual’ activity, but what exactly qualifies as ‘unusual’?
For many customers, using a credit card while on an overseas trip is perfectly normal, yet overseas transactions can often result in a blocked card, turning a relaxing getaway into a stressful experience. Without proper oversight, these systems regularly flag false positives. These outcomes are just as frustrating for customers as they are for staff, who end up spending time resolving complaints instead of focusing on higher-priority tasks like investigating genuine cases of fraud.
Another common source of false positives is static machine learning models. These systems rely on historical data and fail to adapt to changes in consumer behaviour, societal trends, or evolving fraud tactics. As a result, they quickly become outdated, leaving banks vulnerable to emerging threats while continuing to flag legitimate activity as suspicious.
Without ongoing updates, static models not only miss evolving fraud schemes but also create operational inefficiencies that leave staff and customers frustrated. Not only that, but most static models become outdated from the moment they are deployed, due to the lengthy deployment process, meaning the expected performance realised in a proof of concept is never replicated in production.
How to ensure effective fraud detection without sacrificing customer experience
As banking increasingly shifts online, customers are taking full advantage of the flexibility and functionality that online platforms provide. Nationwide, for example, has recently updated its mobile banking app, adding new features that streamline real-time transactions and enhance payment management, giving customers greater control.
However, with this increased flexibility, spending patterns have become more fluid and less predictable, making it harder to distinguish between typical and suspicious financial activity. To ensure fraud detection remains both thorough and non-intrusive, banks must adopt solutions that can adapt to a broader context and continuously evolve with changing behaviours.
Daily adaptive machine learning models (DAM) continuously update based on new user behaviours, the adoption of new financial products, and evolving fraud tactics. This adaptability ensures fraud detection systems stay relevant and effective without disrupting the customer experience.
How does this solution work?
Through our consulting, advisory, and digital services, NTT DATA delivers a robust fraud detection solution powered by Lynx Tech’s advanced technology. This adaptive, daily-updating model enables financial institutions to identify suspicious activity accurately while minimising customer disruption, drawing from data sources such as:
- User devices and locations
- Card and account transactions
- Beneficiary payments
- Incoming funds
- Telephony and branch events
- Authentications and logins
- Onboarding information, and more.
These data sources enable us to construct a detailed, self-learning profile for each user, adjusting daily to reflect changes in financial behaviour and emerging fraud patterns. With this dynamic, adaptive profile, our solution reduces the volume of false positives and allows employees to prioritise high-risk cases, freeing up resources for advanced fraud investigations.
Incorporating constant error checks, data validation, and cyber-defence safeguards, this partnership between NTT Data and Lynx ensures that financial institutions maintain robust security against new and emerging threats. This adaptability and resilience mean banks can stay ahead of potential fraud trends while protecting the customer experience.
Learn more
If you’re interested in reading more about how we can use AI to tackle financial crime, this blog is part of a series in partnership with Lynx Tech. In our previous blog, we explored the threat of money muling and how Lynx’s fraud detection software can help FIs identify muling activity in real time. Read the full blog here.