Incognia launches AI Agent Detection for financial institutions

Incognia has launched AI Agent Detection, a fraud prevention capability designed to help financial institutions assess AI-driven transactions and account activity. The product separates verification of the AI agent itself from approval of the specific action the agent is attempting to perform.

The launch addresses a growing risk as AI agents increasingly make purchases, payments and account changes on behalf of consumers. Incognia argues that confirming an agent is legitimate does not automatically confirm that a customer authorised a particular transaction or account change.

AI Agent Detection evaluates the agent and the action separately

AI Agent Detection combines signed-request verification, agent-origin classification and bot and automation detection with Incognia's broader risk intelligence. When additional assurance is required, financial institutions can compare the request with signals from trusted customer devices, account activity, device integrity, network and location.

André Ferraz, co-founder and CEO of Incognia, said that verifying the origin of a request is different from determining whether the requested action should be permitted. He added that financial institutions need to connect each request to the real customer and the surrounding context while maintaining their existing risk controls.

Web Behavioral Biometrics adds more user signals

Incognia is also expanding its web risk intelligence with Web Behavioral Biometrics. The capability adds signals such as mouse movement, keyboard interaction and copy-and-paste behaviour to existing browser, device, automation and location data.

When further verification is needed, a web session can be linked to a trusted mobile device to provide additional evidence about whether the user behind the interaction is genuine. Incognia combines apartment-level location intelligence with device and behavioural signals to identify trusted users across sessions.

The company provides fraud prevention technology to businesses in financial services, marketplaces and mobility.

MCP Server supports AI-assisted fraud investigations

Incognia has also introduced an MCP Server designed for fraud analysts using compatible AI applications. Analysts can retrieve evidence, identify connections between devices and accounts and organise findings for review using natural-language queries.

The initial version is read-only, meaning AI applications can access and organise investigation data but cannot make enforcement decisions. Final decisions therefore remain under human control.

Financial institutions face new governance requirements around agentic AI

The product launches come as financial institutions increase their use of agentic AI while verification and governance frameworks continue to develop. Jim Mortenson, Fraud Strategic Advisor at Datos Insights, said many firms still lack consistent methods for verifying AI agents, detecting related risks and governing their activity.

He added that investment in control infrastructure at this stage could improve long-term resilience and reduce the need for more expensive redesigns later.