San Jose, Calif. – October 06, 2026 -- Incognia has introduced AI Agent Detection, a new capability designed to help financial institutions distinguish between a legitimate AI agent and an authorized action, as AI agents increasingly make purchases, payments and account changes on behalf of consumers.
Incognia separates agent identity from action risk
The new system combines signed-request verification, agent-origin classification, bot and automation detection and Incognia's broader risk intelligence. When additional confidence is required, financial institutions can cross-check an action against signals from a customer's trusted devices, account activity, device integrity, network and location.
"AI agents change the trust problem for financial institutions," said André Ferraz, co-founder and CEO of Incognia. "Knowing that a request came from a legitimate agent isn't the same as knowing that a specific action should be allowed."
Datos Insights warns FIs lack governance frameworks for agentic risk
Jim Mortenson, Fraud Strategic Advisor at Datos Insights, said many financial institutions are advancing on agentic AI without reliable verification, detection and governance frameworks to manage the underlying risks. He said organizations must invest now in control infrastructure to distinguish between legitimate and malicious humans as well as the agents acting on their behalf, to avoid costly redesigns later.
Web Behavioral Biometrics adds mouse and keystroke signals to risk scoring
Incognia is expanding its web risk intelligence with Web Behavioral Biometrics, incorporating mouse movement, keyboard interaction and copy-and-paste activity alongside existing browser, device, automation and location intelligence. When further verification is needed, the system can connect a web session to a trusted mobile device for additional evidence on whether the person behind the interaction is legitimate.
New MCP Server feeds fraud investigators natural-language evidence retrieval
Incognia's new MCP Server brings its intelligence into compatible AI applications used for fraud investigations. Analysts can use natural-language questions to retrieve evidence, identify connections across devices and accounts, and organize findings for review. The initial MCP experience is read-only, keeping enforcement decisions and actions under human control.