Tracking changes to AI fintech vendors

Incognia launched Real Time Scam Prevention for banks, which flags likely scam payments before money moves by checking for signs of a real world relationship between sender and recipient, drawn from apartment level location intelligence and persistent device recognition. It combines that with other risk signals and pools intelligence across participating banks without any of them seeing another's location data. Incognia reports identifying about 70% of scam transactions in an analysis at a bank with more than 80 million active users.

Our read

Authorized push payment scams pass device and login checks because the real customer sends the money, so a sender to recipient relationship signal gives banks a new input for stopping payments. The 70% figure comes from an unnamed bank with no published method, so buyers should ask for validation detail.

Incognia launched AI Agent Detection for financial institutions, which identifies agentic activity and assesses the risk of each action an agent initiates separately from the agent itself, combining signed request verification, agent origin classification and bot detection with Incognia's device, location and account signals. When an action needs more confidence, it can be checked against the customer's trusted devices and account history. The same release adds a read only MCP server that lets fraud analysts query evidence and device and account links from compatible AI apps.

Our read

As customers start sending AI assistants to act on their accounts, fraud teams need to judge the action as well as the agent, and this lets them do it with signals they already trust. Incognia does not state GA status or customers; the web behavioral biometrics in the same release were already logged on 23 September.

Incognia added behavioral biometrics to its web risk intelligence, so it can now assess mouse movement and clicks, keyboard interaction, and copy, paste and cut activity, including whether those events are trusted browser events or synthetically generated. The signals are evaluated alongside its browser, device, integrity, bot and automation, and Verified IP Location signals, with step up to Cross Device Authentication on a trusted phone when web evidence is not enough.

Our read

Banks and fintechs defending web logins and account changes get a signal for scripted or synthetic sessions without adding a challenge for every user. The post states the business, not Incognia, makes the final risk and authorization decision.

Incognia launched AI Powered Browser ID, a web risk product that applies a transformer architecture to the semantic relationship between more than 200 browser metadata signals. Rather than hashing attributes into a static fingerprint, it tokenizes the metadata so an identity stays stable when the technical environment changes.

Our read

Fraud teams keep recognizing a returning user through routine browser updates and configuration changes that break conventional fingerprinting. That removes a standing source of false positives without loosening the rules that catch real web fraud.