AI agents report incorrect banking fees and security risks

AI agents utilizing open finance APIs are reporting incorrect banking fees and exhibiting security vulnerabilities. During a panel at Money20/20 USA in Las Vegas, experts highlighted that these agents often provide inaccurate information to customers, such as claiming services are free or providing random fee amounts.

Beyond financial inaccuracies, testing by APIContext revealed significant security risks. In some instances, AI agents failed to refuse requests for account balances even when presented with false power of attorney. Additionally, there are concerns regarding how agents interpret available balances, which could lead to unintended overdrafts or unauthorized data access.