Thanks to artificial intelligence (AI), fraud has become increasingly complex.
Fraud can be done in the form of Deepfake videos or sounds, and AI produces clones of family relatives who are said to be in emergency situations and requires immediate cash transfers. AI can write more compelling phishing emails, removing obvious signs like broken English. According to a FBI report, AI can also forge such as driver's licenses and fraudulent driver's licenses, such as a FBI report.
“Fraud only gets worse with the creation of generative AI,” said Mike de Vere, CEO of Zest AI.
According to a March 2025 report by the Federal Trade Commission (FTC), the losses caused by fraud reached $12.5 billion in 2024, a 25% increase from the previous year. More and more people are also reporting that they lost money due to fraud: 38% last year, compared with 27% in 2023.
Investment scams have cost people the most money, totaling $5.7 billion, up 24% from the previous year. The second highest was the impostor scam, at $2.95 billion. However, impostor scams are the most common fraud, followed by online shopping fraud.
It is worth noting that consumers lose more money through bank transfers or cryptocurrencies, than all other payment methods combined, the FTC said.
According to PYMNTS intelligence research, in partnership with I2C, 28% of consumers became victims of credit card fraud last year. Additionally, according to the Consumer Credit Economy: Credit Card Fraud, 37% said they were “very” or “extremely” worried about becoming a victim of such fraud.
He added that in an interview with PYMNTS, he said that fraud losses are expected to reach $40 billion by 2027. Fraud tools are becoming increasingly accessible, he added that for only $20, criminals can do things like creating fake IDs and Pay Stubs.
Read more: 37% of consumers are highly concerned about credit card fraud
What financial institution believes it wrongly
Based on his experience working with banks and credit unions, De Vere shares his insights into five myths about fraud prevention that can make organizations vulnerable.
Misunderstanding 1: Small banks can safely resist fraud
The first misunderstanding is that fraudsters target only major financial institutions. In fact, among the 10 banks and credit unions, including smaller bank unions, reported fraud losses of more than $500,000 last year.
“This has a disproportionate impact on smaller financial institutions,” Deville said. “The fraudsters who are opposite to the Citi IT team may be more than [targeting] A small credit union outsources it to them. ”
Misunderstanding 2: Transaction monitoring is enough
Many institutions believe that monitoring individual transactions provides adequate fraud prevention protection. For example, look at the customer's credit card model to find out if there are fraudulent purchases.
But, Deville said, this narrow approach misses the wider pattern of behavior that AI can detect. He shares this real-world example: The fraudster opens a credit card at a credit union, charges about $100 a month, and pays it off regularly. In itself, this behavior does not cause red flags. However, the criminal did the same thing at several credit unions. The individual eventually applied for and received a personal loan, obtained a credit card to the greatest extent, and disappeared with the money.
Misunderstanding 3: Safety requires friction
The third myth revolves around the idea that to ensure security, financial institutions must go through several baskets, such as asking for answers to solve security issues, etc. It creates friction in the customer experience. These binary fraud systems – are they fraud? Yes or No – may cause problems unnecessarily.
He shared his personal experience of being marked as ID fraud in a car loan application simply because his last name was squeezed together. “A AI solution could have looked at my credit report and saw…my two credit cards actually smashed my last name together, so I might not be a fraudster.”
Misunderstanding 4: Manual Comments Capture Fraud
Humans should be the gold standard when it comes to fraud, but Deville believes that their experience is only as good as that. In addition, manual review is limited by the reviewer’s experience in the institution.
Instead, AI models can consume trillions of data points to identify fraud patterns. “This is far beyond humanity,” Deville said.
Misunderstanding 5: All fraud solutions are equal
The final myth is that fraud prevention solutions are interchangeable. Deville said many of the available solutions are incomplete and creates blind spots within security coverage.
He said a powerful fraud prevention solution should provide probability scores rather than binary “fraud/free” decisions, trained in a comprehensive dataset and tailored to the needs and geography of the organization. This approach allows organizations to identify local fraud rings and take appropriate security measures.
De Vere advocates a collaborative approach to fraudulent collaborative approach: “We need to think less about it being a competitive issue, but more of a collaboration issue.”
To this end, enthusiastic AI created a consortium to share fraud experiences, enabling AI models to learn from attacks on one institution to protect others in the same ecosystem.
