80% of banks had attacks from AI agents, a survey finds

BioCatch makes software that finds fraud from how persons type and use a mouse. Its survey of 1,440 bank staff in 25 countries found this. 80% tell that their bank had attacks from AI agents. An AI agent is software that does tasks for a user. BioCatch also released AiRE, a tool that finds AI agents on accounts before money moves.

How it works

BioCatch tells that AiRE finds AI agents that work for account holders or for criminals. It looks for unusual behavior, such as synthetic inputs and browser data. Each bank gets a risk signal.

The numbers

• 84% of respondents see AI agents as the primary weak point for attacks next year.
• 72% think it is not easy to tell safe agent steps from dangerous steps.
• 9 of 10 brokers let Claude connect to customer accounts, an FM Intelligence analysis found.

The source gives no test results for AiRE.

What is not known

BioCatch tells that the effect of the tool depends on how a bank uses it. It is not clear if banks will stop all agents, let only some in, or let all in. Reports tell that UK banks, such as Barclays, stop agents. Robinhood tells that only the customers have the liability for each task that their agents do.

Insight: Banks stop agents but brokers let them in

Reports tell that UK banks stop AI agents. Large US banks limit them. Brokers let agents in and put the liability on the customer. This is a divide on who has the cost of an incorrect agent task, in our reading. A company that makes tools to find agents gets more control when each bank must check each session. Look at which agents banks let in first.