September 2, 2026
The AI Agent Paradox in Brazil
Brazil leads AI agent adoption, but 80% of companies have already reversed implementations due to governance issues. Data leaks took down 39% of those projects.
Brazil runs more AI agents in production than the United States. 76% of Brazilian companies already have agents operating for real, compared to 67% of American ones and 62% of the global average. 80% of the Brazilian companies surveyed have already stopped or reversed an AI implementation over governance issues. Data or personal information leaks brought down 39% of those projects.
The data comes from a global study by Sinch, a communication infrastructure provider that also sells AI agents for customer service, with more than 2,500 executives surveyed across ten countries. In Latin America, 41% of companies shut agents down after a leak. In North America, 26%.
Reversal is not unique to Brazil. Globally, 74% of companies went through the same experience. The proportion tied to leaks varies by country: 39% in Brazil, versus 24% in the United States and 16% in Canada. Did whoever approved these agents define, before launch, the criterion for taking them offline? The sources do not say. That is what separates two kinds of decision: the company that defines the rollback criterion alongside the approval scope, and the one that only defines it after the leak.
66% of Brazilian companies plan to increase AI investment by more than 25%, with no guarantee that the leak rate changed the design of the approval process. A rollback metric builds trust when it enters as a criterion in the agent's approval, not as an explanation after the previous one leaked.
Sources: Olhar Digital · Sinch (The AI Production Paradox) · TI Inside