Security

Central Bank Sees ‘Intensive Use’ of AI Agents in Cyberattacks on Brazil’s Financial System

· 6 min read · Security

Central Bank Sees ‘Intensive Use’ of AI Agents in Cyberattacks on Brazil’s Financial System

Brazil’s Central Bank has seen criminals make intensive use of specialised AI agents in cyberattacks on the financial system since October, a senior supervisor said on November 17, according to Valor Econômico. Aristides Cavalcante, head of the bank’s department of strategic management and specialised supervision, told Febraban’s anti-money-laundering and risk-management congresses in São Paulo that the risk level for Brazilian institutions has risen significantly and is no longer where it stood at the start of the year.

Cavalcante said the regulator has recorded 115 cyber incidents in 2026, of which 78 were classed as fraud, meaning attacks aimed at taking money from institutions rather than scams against their customers. In one case this year, which he presented without naming the firm, AI agents were used at every stage of an attack on a virtual-asset service provider. Organised crime has not left Brazil, he argued. Bank robberies have faded from the news because gangs moved from the physical world to the digital one, and they now go after the technology infrastructure of financial institutions as well as their clients.

What it means in practice

Central Bank Sees ‘Intensive Use’ of AI Agents in Cyberattacks on Brazil’s Financial System — contextual photo

The bank’s survey of 1,365 institutions points to basic gaps that AI-assisted attackers can exploit faster. Some 43% said they have no mechanism to protect internal networks against unauthorised devices, 31% do not use multi-factor authentication in critical processes and 25% do not regularly review staff access. About 20% have no policy for managing third-party suppliers, and only half of the 770 institutions that offer APIs check them regularly for tampering. Cavalcante listed measures taken since 2025, including the end of pooled accounts, new minimum capital rules, Banking as a Service regulation and an emergency tool that lets institutions halt settlements, which went into production this month.

The warning contrasts with the sector’s own push into AI, from Banco do Brasil’s gradual autonomy for AI agents to Google Cloud’s cyber agents for Brazilian clients, and echoes the CVM chief’s view that only AI can fight AI. What to watch is whether the Central Bank tightens its minimum cybersecurity controls, how quickly institutions close the gaps in its survey, and whether it starts publishing figures on incidents driven by AI agents.

← Back to AI4Brazil