Fraud is evolving at an unprecedented pace. Looking ahead, what do you believe will be the defining fraud challenge for financial institutions over the next three to five years?
Today’s social engineering-based attacks still have one physical bottleneck. A convincing scam call requires people to be recruited, trained, scripted, supervised and to sit in a building with telephony infrastructure that looks like a legitimate business. That is the most expensive and most fragile part of the operation, and it is what limits how fast a scam campaign can scale.
Deepfake voice is already becoming convincing enough to remove one barrier from scam operations. The remaining gap is autonomy: systems that can hold a full conversation, handle objections and adapt in real time without a human on the line. Once that gap closes, the call-center layer becomes optional. A one-person operation could produce the output of a large fraud floor, without recruitment, training, supervision or premises.
At that scale, every call becomes part of a feedback loop. Scripts can be tuned automatically based on what works: which objection lowers resistance, which pretext holds attention longest and which victim profile converts. The conversation itself becomes self-optimizing in real time for each target. I think this will soon cease to be science fiction, and within five years, we are likely to see cases proving it is a new operating reality for fraud.
Many fraud strategies focus on responding to emerging threats. How can organisations become more proactive in anticipating what’s coming next rather than reacting to the latest attack?
Reactive strategies fail because they are built around known attack signatures. Anticipation comes from monitoring behavior around the transaction rather than the attack pattern itself, because the underlying human and technical signals change far more slowly than the scenarios built on top of them.
Social engineering, remote access abuse, and account takeover leave behavioral traces – hesitation, dictated input, unusual navigation – before money moves. A model built on behavioral, device, and network context can catch scenarios it has never seen.
Most controls wake up at the payment. By then, the reconnaissance, enrolment, and beneficiary change have already happened. Continuous monitoring from login to logout gives you early indicators and time to act.
Additionally, mule activity sits behind almost every fraud typology. Identifying mule accounts, both inside and outside the bank’s customer base, disrupts cash-out infrastructure before the next campaign can use it. Cross-institution patterns and analyst feedback must flow back into detection models on a short cycle. A team seeing only its own cases stays one campaign behind.
AI is reshaping both sides of the fight against fraud. Where do you see the greatest opportunity for financial institutions to use AI to strengthen fraud prevention and detection?
Attackers are using AI to scale activities that used to be manual: deepfake voices, localized scam scripts, synthetic identities and automated account testing. Both volume and quality increase, so defenses built around human review capacity cannot scale with the threat.
In my opinion, the biggest opportunity for banks lies not in a better transaction-scoring model. It is behavioral intelligence at the session level.
AI models trained on behavioral biometrics and session behavior can detect when a legitimate user is acting under coercion or instruction. Hesitation, dictated input and unfamiliar navigation can signal that the customer is real, but the behavior is shaped by external pressure.
How important are people, first-line training and organisational culture in building an effective defence against increasingly sophisticated fraud?
I think that technology sets the ceiling for what you can detect, but people decide how much of that detection actually turns into prevented loss.
The first line is where detection becomes intervention – a model can flag a session as likely social engineering, but a contact center agent is the one who has to break the spell with a customer who is actively being coached by a fraudster to de-manipulate them. That conversation is a skill, and it needs training in scam psychology – not just on process. People are still irreplaceable in this discipline, and I think it will be for years to come.
What session are you most looking forward to seeing at the Fraud FS Summit 2026 and why?
For me, the session I am most looking forward to is “Transaction Forensics – Tracing Mule Networks Across Institutions”. I think that tracing mules will become one of the most important controls banks can introduce.
In my opinion, cash-out infrastructure is the most efficient point of disruption available to banks and one of the least exploited.
What interests me specifically is the cross-institution dimension. Most banks can identify suspicious inbound accounts inside their own customer base, but very few can follow the chain once funds leave.
About Lukáš Jakubíček
Lukáš has over 20 years of experience in IT and, since 2018, has been part of the security experts at ThreatMark. His work has given him deep insights into cybersecurity and a thorough understanding of the technologies exploited by online fraudsters.
As a seasoned speaker, he regularly presents at domestic and international cybersecurity conferences and delivers training to fraud analysts worldwide. Daily collaboration with cyber experts keeps him at the cutting edge of emerging threats and effective prevention strategies.
Pre-Event insights from ThreatMark – Associate Partner of Fraud FS Summit
The Fraud FS Summit brings together senior fraud and financial crime professionals for a full day of presentations, panel discussions, interactive roundtables and networking; inclusive of 2 course lunch and drinks reception with canapés. Join the UK’s leading fraud conference in London, view the latest agenda for the Fraud FS Summit here.
