Itaú hires psychologists to train AI against scam 'trances'

Itaú Unibanco has hired a team of psychologists to train its conversational artificial intelligence assistant to help pull fraud victims out of psychological manipulation. Felipe Tambelini, the bank's fraud prevention director, stated that Brazil is experiencing an epidemic of scams that is out of control, driven by social engineering schemes such as fake sales, fake customer service centers, and impersonation of lawyers.

The lender's system currently identifies when a customer opens the mobile app while on a phone call and generates warning screens. The application features nearly 40 different types of contextual alerts, which are customized using artificial intelligence based on the user's demographic profile and the specific type of scheme detected.

Because many users continue with transactions even after reading alerts, visiting a branch, or signing liability statements, Itaú is shifting toward an interactive conversational assistant. Instead of simple affirmative or negative buttons, the AI will conduct dialogue using questions developed with psychoanalysts. The bank has also brought actual scam victims to a mirror room at its headquarters, where specialists evaluate their accounts to pinpoint behavioural triggers.

Tambelini noted that deepfake scams remain limited due to the high computational cost of tokens for criminals, who find it cheaper to target user psychology. In parallel, Itaú expanded its Protected Mode on its Superapp, adding a security layer that protects investment wallet balances and information when devices are outside registered secure Wi-Fi networks, requiring identity verification for access.