AI in financial crime prevention: what works, what’s risky, and what’s hype
MC
Mirela Ciobanu
26 Mar 2026 / 5 Min Read
Ivan Stefanov, CEO and Co-founder of NOTO, breaks down the reality of AI in fraud and AML; where it delivers measurable value today, where it introduces risk, and how financial institutions should approach adoption responsibly.
Artificial intelligence has become the defining buzzword in financial services, but in fraud prevention and AML, the stakes are too high for hype-driven adoption. In this Leadership Insights conversation, Ivan Stefanov, CEO and Co-founder of NOTO, offers a grounded, practitioner-led perspective on what AI means for financial crime teams today.
Clarifying the AI terminology
One of the first points Ivan makes is conceptual clarity. AI and machine learning are not interchangeable. Machine learning is a subset of artificial intelligence, an instrument within the broader AI discipline. The current wave of interest, however, goes beyond traditional ML models that banks have used for years in transaction scoring.
Large language models (LLMs), generative AI, and agentic AI introduce new capabilities. Generative AI can summarise complex case data, explain why rules were triggered, or draft client communications. Agentic AI can act on those outputs: escalating alerts, requesting documentation, or initiating workflows. But this does not mean they are suited for real-time transaction scoring at scale. The true opportunity lies elsewhere.
Why AI is dominating the financial crime agenda
According to Ivan, the surge in AI adoption is driven by a mix of real performance improvements, competitive pressure, regulatory expectations, and plain FOMO. Financial institutions are observing early adopters reporting positive ROI and feel pressure to keep pace.
At the same time, criminals are already leveraging AI aggressively. Automated phishing campaigns, identity spoofing, and scalable scams are increasingly powered by AI tools. Unlike regulated institutions, criminals are unconcerned with hallucinations or bias if even a small percentage of attempts succeed, the model works for them. This dual dynamic (offensive use by criminals and defensive adoption by banks) makes AI impossible to ignore.
Where AI adds measurable value today
Ivan structures financial crime prevention into three stages:
Real-time transaction scoring and decisioning
Review and investigation
Analytics and feedback loops
While LLMs are not yet capable of handling high-speed real-time scoring under strict latency requirements, they can significantly augment the second and third stages.
AI proves most effective in:
Alert triage and case summarisation
Aggregating data from multiple sources
Supporting investigators with contextual insights
Monitoring dashboards and identifying patterns
Enhancing productivity in high-volume environments
The dominant model emerging today is augmentation, not replacement. AI increases SME accuracy and efficiency rather than removing human oversight.
Governance, risk, and hallucinations
The risks, however, are real. Hallucinations (outputs that are plausible but incorrect) pose a significant threat in sanctions screening and AML investigations. A convincing but inaccurate summary could result in a missed sanctioned entity, creating regulatory exposure.
Ivan emphasises several guardrails:
Clearly defined use cases
Strong data governance and GDPR awareness
Compliance with frameworks such as the EU AI Act
Operational resilience planning
Human-in-the-loop validation
Completely removing investigators from the loop risks knowledge erosion. If institutions rely solely on AI-generated outputs without maintaining internal expertise, they create systemic vulnerability.
Looking toward 2030: what leaders will regret (or celebrate)
When asked to fast-forward to 2030, Ivan suggests that leaders who went ‘all in’ on AI as a silver bullet may regret abandoning disciplined implementation processes. Decisions driven by hype rather than structured evaluation could lead to costly missteps.
Conversely, institutions that:
Conducted proper scoping and evaluation
Resisted internal political pressure
Maintained governance discipline
Integrated AI gradually and strategically
will likely look back with confidence.
Ivan’s overarching message is one of measured optimism. AI is neither a bubble to ignore nor a magic solution to embrace blindly. Its strongest potential lies in becoming a reliable ‘second pair of eyes’, a digital partner that challenges and enhances human expertise rather than replacing it.
We invite you to learn more about this topic by watching our conversation.
About author
Ivan Stefanov is the CEO and Co-founder of NOTO, with extensive experience in fraud prevention across financial services and the crypto industry. He previously held senior risk and fraud leadership roles at Groupon, Paysafe Group, and Crypto.com.
The Paypers is a global hub for market insights, real-time news, expert interviews, and in-depth analyses and resources across payments, fintech, and the digital economy. We deliver reports, webinars, and commentary on key topics, including regulation, real-time payments, cross-border payments and ecommerce, digital identity, payment innovation and infrastructure, Open Banking, Embedded Finance, crypto, fraud and financial crime prevention, and more – all developed in collaboration with industry experts and leaders.
