A in payments: it no longer just detects fraud, it now learns how you shop to identify when it wasn't you

The growth of digital payments has been accompanied by a silent transformation of financial fraud. What once relied mainly on credential theft or card cloning now incorporates social engineering, synthetic identities, and artificial intelligence tools capable of making attacks more convincing and scalable. In this scenario, banks, fintechs, and companies face an increasingly complex challenge: protecting users without compromising the speed and simplicity that characterize digital payments.

According to a survey by the Datafolha Institute in partnership with the Brazilian Forum for Public Security (FBSP), approximately 24 million Brazilians were victims of financial scams involving Pix, credit cards, and payment slips between July 2024 and July 2025, accumulating estimated losses of nearly R$29 billion. Furthermore, information from the National Strategy to Combat Corruption and Money Laundering (ENCCLA) indicates that more than half of Brazilians have already suffered some type of financial fraud.

Today, three trends are redefining risk management in the financial sector. On one hand, fraud is growing rapidly and has changed in nature, driven by social engineering and the use of generative artificial intelligence. On the other hand, regulation demands real-time, transparent, and auditable monitoring systems. And finally, defense is becoming increasingly dependent on artificial intelligence and low-friction authentication mechanisms capable of protecting the user without compromising the payment experience.

Imagine a consumer who buys coffee every morning, at the same time and in the same coffee shop. Suddenly, a few minutes later, a high-value purchase appears, made on a different device and in a different city. At that moment, the most important question is no longer "what is the password?", but "does this behavior make sense for this customer?". This is the new frontier of fraud prevention: instead of relying solely on passwords or credentials, artificial intelligence learns how each person shops to identify when the person paying might not be them.

The industry's response points to models capable of interpreting behaviors, not just applying predefined rules. Purchase frequency, usual values, times, location, devices used, and types of establishments are some of the signals that allow for the construction of dynamic consumer risk profiles. When a transaction deviates significantly from this pattern, the systems can identify anomalies and assess the level of risk before authorizing an operation.

This change also responds to an economic need. In addition to combating fraud, institutions need to reduce so-called false positives, legitimate transactions rejected due to suspected risk. For banks, fintechs, and companies, the challenge is not only to block fraudulent activities, but to do so without generating unnecessary friction or impacting the conversion of legitimate payments.

Pressure on the ecosystem is expected to increase in the coming years. A study by GlobalData estimates that losses associated with scams via Pix could reach approximately R$11 billion per year by 2028, while data from Serasa Experian shows that, in 2025, there was an attempted fraud every two minutes in Brazilian commerce. In this scenario, risk management will cease to be merely a protective function and will become a strategic element of the digital financial experience.

*Rodrigo Rodrigues, COO and co-founder of Akua