While many have concerns about using artificial intelligence in finance, data from Feedzai’s report The AI Shift: Transforming AML Compliance into Competitive Advantage shows many are using it to great effect. The report also counters some misconceptions about AI usage.
The usage numbers are quite strong:
- 71% of AML professionals say their organization already uses artificial intelligence/ML to fight fraud and financial crime;
- 96% say regulators accept/encourage artificial intelligence adoption (with 65% “fully accepting”);
- 95% say model explainability and transparency are must-have requirements;
- 62% report more than a 40% reduction in false positives after adopting AI; and
- 66% report great than 40% improvement in team efficiency.
Where artificial intelligence is being used, and how it helps
Artificial intelligence in AML is commonly deployed in four ways. Supervised machine learning sees systems trained on labelled examples like past data. This assists with pattern detection and alert prioritization.
Unsupervised machine learning eschews labelled data, instead opting for analyzing customer behavior in search of anomalies and unusual activity. Generative AI systems can draft summaries, collect external information and highlight details. It frees analysts for higher-level activities.
Agentic AI models work independently on specific actions and goals. They may take actions depending on preset parameters.
Artificial intelligence improves staff efficiency by identifying priority alerts and analyzing SAR filings to help analysts learn common characteristics that contribute to escalation and further action. Labels can be adjusted as priorities change.
Public information like enforcement actions and regulatory guidance is scanned for emerging trends. Artificial intelligence can also be used to create internal knowledge bases that learn from successful investigations.
The end result is staff are freed up to do more meaningful and high-value tasks.
Various artificial intelligence enforcement approaches leave exploitable gaps
Feedzai’s head of industry relations Robert Harris said different countries take different approaches to regulation. Some, like the United Kingdom, have forced banks to compensate fraud victims. Others take a cross-sector strategy by focusing on other sectors that contribute to fraudulent behavior like telcos and big tech.
While each one has its merits, it leaves gaps as AI-aided fraud is a global activity. Criminals leverage jurisdictional gaps to maximize their activities. More effort need to be directed toward harmonizing regulatory action worldwide.
“What we find is a: regulation is too slow; b: too contradictory in some areas; and c: never solves the problem, especially in things like AI,” Harris said. “I don’t think you can ever regulate AI; it just moves too fast. As soon as you put regulations in place, the criminals will move into new areas first.”
Regulation must determine who is accountable should some artificial intelligence-aided process run errant. Harris said every bank uses multiple such systems. How do they know what is built into them?
Many vendors also look to solve their needs before the customers. Harris argues for the opposite, saying that ultimately, systems should be measured on how they impact the consumer not whether they comply with regulations. Feedzai’s systems offer a balance of fairness, responsibility and the ability to reach business objectives.
How to create a unified global artificial intelligence approach
While financial institutions are using artificial intelligence for more tasks, Harris said concerns about regulation have slowed adoption rates. Institutions must brace for the worst, and if they are uncertain about how artificial intelligence might be regulated, they may sit it out until the smoke clears.
Harris counters that there are enough regulations. More clarity is what is most needed.
How do regulators from around the world unify their approach? Harris said a promising strategy is to focus on stopping popular illegal money corridors. Instead of boiling the ocean, focus limited resources on where they’ll have the most impact.
Crime evolves. For example, whole cities in Asia are focused on committing financial crime. The United States and China are among those trying to stop them.
“In our space over the last five years, crime has become far more organized at a national level,” Harris said.
