Overview
- BioCatch’s 2026 survey of 1,440 fraud, AML and risk professionals at banks in 25 countries found that 80% reported encountering attacks that used autonomous or agentic AI to access accounts or probe systems.
- The company announced AiRE, a detection capability that it says identifies agent-like interaction patterns—such as synthetic inputs, unusual browser permission profiles and agent infrastructure signals—before funds are moved.
- In response to reported agent activity and worries about disabling of internal monitors, many banks have tightened or blocked third-party agent access and shifted toward read-only or vetted connectors to limit exposure.
- Some retail brokers have enabled agent integrations for trading, and firms such as Robinhood require a separate Agentic Account while placing responsibility for agent construction and actions on customers.
- Experts and firms warn that distinguishing legitimate AI-assisted users from malicious agents is hard, which has prompted calls for stronger authentication, rotating secrets, longer logs and greater regulatory scrutiny to clarify liability and reporting.