Skip to content

Press Release

ThreatFabric launches Agentic AI Risk and Know Your Agent (KYA) to distinguish trusted AI agents from malicious ones inside banking apps

FOR IMMEDIATE RELEASE

 

New Fraud Risk Suite capabilities combine a New AI Model, Device Risk, Behavioural Risk and Mobile Threat Intelligence to identify AI agents and assess their intent across the banking session, from onboarding, login to payment.

Amsterdam, 19 November 2026 – ThreatFabric today announced Agentic AI Risk and Know Your Agent (KYA). A new AI model and new capabilities within Fraud Risk Suite help banks distinguish legitimate AI agents from malicious, hijacked or fraudster-controlled ones.

Rather than trusting what an agent claims to be, KYA looks at what is harder to disguise: the device it runs on, the session it creates and the way it behaves. Agentic AI Risk brings these signals together across the entire mobile and web banking session, helping banks determine whether an agent is acting for the customer or against them, before a payment is made.

 

Agents are moving inside the banking app

The launch comes as Google opens Android further to AI agents. AppFunctions has moved to beta, while the Computer Control framework provides a fallback that enables assistants such as Gemini to operate apps with the customer's approval, even without direct integration by the bank.

This represents an important change for financial institutions. AI agents are moving beyond the edges of the customer journey and into the banking session itself, navigating apps and potentially initiating transactions on behalf of customers.

Criminals are moving quickly too. AI lowers the barrier to sophisticated fraud and gives criminals new ways to automate, scale and orchestrate attacks. Agentic AI Risk is designed for the point where these two developments meet.

 

Know Your Agent: device, session and behaviour

An AI agent can claim to be anything. KYA starts with the signals it cannot easily hide and combines three layers already available within ThreatFabric's Fraud Risk Suite:

  • Behavioural Risk includes a new AI model built on Agentic Interactions. It uses ThreatFabric's patented behavioural identity technology to identify whether a session is being driven by the genuine customer or by automation. In a reference deployment, every agent-generated payment was identified as “not the customer.”
  • Device Risk detects automation and manipulation on the device, including Accessibility Service abuse already used to control banking apps in real-world fraud.
  • Mobile Threat Intelligence, built on more than a decade of research into mobile banking malware, continuously identifies emerging techniques and agentic attack patterns.

Agentic AI Risk combines these signals across the entire session, assessing login, navigation and payment as a connected journey rather than a series of isolated checks. This enables banks to move beyond the question “Is this an agent?” to the questions that matter more: Whose agent is it, and what is it trying to do?

 

Every path an agent takes into the app

Agentic AI Risk is designed to provide visibility regardless of how an AI agent interacts with the banking application:

    • AppFunctions, where banks or app developers expose specific functions that an AI agent can invoke.
    • Computer Control, where an AI assistant operates the application through its user interface without requiring a dedicated agent integration.
    • Accessibility-based automation, a route already being exploited to manipulate banking apps today.

ThreatFabric expects increasingly capable computer-control technology to create a new surface for automated and assisted fraud, following a pattern previously seen with remote-access tools. Technologies built for legitimate purposes can quickly become useful to criminals once they provide a scalable way to interact with a victim's device.

 

Coverage across 22 agentic fraud use cases

ThreatFabric mapped its Fraud Risk Suite capabilities against 22 agentic fraud use cases. Device Risk, Behavioural Risk and Mobile Threat Intelligence provide strong, high-confidence signals across 90% of use cases, spanning five areas:

    • Mobile and on-device threats: Accessibility Service misuse, AI-enabled malware and device takeover, and AI assistants operating banking apps.
    • Agent detection in the session: humans versus agents and bots, agents impersonating human behaviour, agentic browsers, cloud-hosted agents and agent farms.
    • Agent-driven scams: victim coaching, AI voice agents active during banking sessions, and hybrid human-agent attacks.
    • Customer lifecycle: agentic account takeover, automated account opening at scale.
    • Trust and intent: legitimate versus malicious agents, legitimate versus malicious intent, and hijacked agents.

The precursor to agentic fraud is already here

The transition is not theoretical. ThreatFabric has already observed Android applications using accessibility capabilities to automate actions and transfers inside banking apps. These techniques provide a preview of how agentic technology can change mobile fraud. As legitimate AI assistants gain greater ability to operate applications on behalf of consumers, similar capabilities can be abused by malware, fraudsters or compromised agents. This makes simply detecting an AI agent insufficient.

 

Banks need to understand the intent behind it.

With the new AI model, combined with device, session and behavioural intelligence, ThreatFabric aims to help banks distinguish trusted automation from malicious activity rather than blocking AI agents outright. This allows financial institutions to embrace legitimate agentic banking while maintaining control over the risks it creates.

“Remote-access tools were never designed for fraud, yet they became one of the scammer’s favourite instruments. AI agents create a similar challenge. Banks need to know not only that an agent is operating their app, but whether it is acting for the customer or against them. That distinction is what will allow banks to embrace agentic banking without opening the door to agentic fraud,” said Han Sahin, Co-founder and CEO of ThreatFabric.

 

About ThreatFabric

ThreatFabric delivers fraud prevention built on leading threat intelligence. Its Fraud Risk Suite provides real-time fraud risk signals across mobile and web banking, while its Mobile Threat Intelligence gives financial institutions insight into evolving criminal tactics and techniques.

ThreatFabric protects dozens of millions of online banking customers worldwide and leads global Android banking malware family research. Learn more at threatfabric.com.