Banks use smart AI as fraud costs reach $41.69 billion by 2028
New AI called Graph Neural Networks stops credit card fraud better than old methods. Banks face $33.79 billion in fraud losses in 2025. The new AI looks at how payments connect to each other. Old systems miss these links.
Big payment companies now use this AI. Tests show it cuts false alerts by 200%. It also works 300% faster. NVIDIA made a new system that mixes both old and new methods. This gives good results that banks can explain.
Old Systems Have Big Problems
Old computer systems have run fraud checks for years. These include decision trees and other tools. They look at each payment by itself. They check things like payment size and store type to spot bad activity.
“Old systems find patterns in payment data,” said Brad Rees from NVIDIA. “But they miss how things connect to each other. New AI looks at how payments link together. It doesn’t just check each payment alone.”
Fraud Costs Keep Going Up
This change shows fraud is getting smarter. So is the tech to stop it. Online stores will lose $33.79 billion to fraud in 2025. This goes up to $41.69 billion by 2028. Fraud cases will hit 337 million by 2026.
Graph Neural Networks treat payments, stores and buyers as dots on a map. The AI looks at how these dots connect. It finds fraud rings and money washing that old systems miss.
Tests Prove New AI Works Better
New research shows this AI handles money networks better. A study of 33 research papers found key results. The new AI beat old systems at finding fraud. It’s really good at finding hidden bad guys and smart attacks.
The tech fixes big problems with old ways. Old systems think each payment stands alone. This fails to catch organized fraud. New AI shares info between connected payments. This finds fraud patterns across many payments and people.
How the New AI Works
“The system drops some messages between connected dots,” researchers said in a fraud study. “This keeps good and bad signals clear. It stops bad guy profiles from getting mixed up with good messages.”
But old systems still work well in some cases. Experts say decision trees remain the “go-to tool” for organized data. They train faster and cost less to run.
Business Leaders Love New AI
Nate Matherson from Numeral talks about AI’s power. “New AI models are changing how we fight fraud,” he said. “They give very accurate fraud detection and real-time analysis.”
Both techs face setup challenges. New AI needs lots of computer power and graph knowledge. Old systems struggle with changing fraud and too many false alerts.
Mixed Approach Works Best
The mixed approach looks most promising. NVIDIA’s system combines new AI with old tools. This gives better accuracy while staying clear. Banks need this to follow rules.
Companies adopt this faster across payment networks. Uber, Google and Pinterest use new AI systems now. Small banks still use old methods due to limited tech resources.
Clear Results Drive Growth
AI-powered fraud tools report 13% better payment clarity. They also show 30-40% more people using prevention tools. The tech can process huge payment networks in real-time. This makes it very valuable for big fraud prevention.
What Comes Next
Experts predict both approaches will keep merging. As computer costs drop and AI tools get easier to use, the new tech will slowly replace old methods. This will happen in high-volume places. Old methods will still work for cost-focused uses.
The future looks bright for stopping fraud. New AI gives banks powerful tools to fight smart criminals. As more companies adopt these systems, fraud losses should start dropping. This saves money for businesses and buyers alike.
Banks that move fast to adopt new AI will have big advantages. They’ll catch more fraud while bothering fewer good customers. This leads to happier customers and better profits.
