Stopping fraud without stopping customers’ money

A customer who usually pays by card in physical shops may suddenly buy an expensive airline ticket online. For a bank, the challenge is to decide in a fraction of a second whether the purchase is legitimate or fraudulent.

Artificial intelligence has made that challenge harder. Fraudsters can now produce convincing phishing messages, fake advertisements, deepfake videos and voices that imitate bank consultants or family members. Preparing such material has become faster and easier, while the cost of cybercrime is growing three times faster than global GDP. In 2024, fraudulent bank transfers cost consumers and companies 2.5 billion euros in Europe, according to the figures cited in the article.

The same technology is also being used to defend payment systems. Featurespace, a company owned by Visa, focuses on how individual customers normally behave rather than relying only on known fraud patterns. Its adaptive behavioral analysis and automated deep-learning networks look for activity that does not fit a customer’s transaction profile.

Featurespace technology, combined with data from Visa’s network, supports Visa A2A Protect. The system identifies fraud, account takeovers and coordinated criminal activity in real time, including patterns that one bank might miss on its own. One of the largest European banks using the service cut fraud by more than 50% and unnecessary alerts by over 40%.

The technology is also applied during transaction authorization. Adaptive AI models analyze payments in real time for card issuers and payment processors, helping them identify risk, improve approval rates and reduce false declines. That allows financial institutions to expand payment services while protecting customers’ money.