Zimperium zLabs Reveals Android Trojan RatHat Abusing AI Assistant to Operate Devices

On September 18, 2026, Zimperium's zLabs research team disclosed a new Android trojan named RatHat. The malware primarily targets users in Chinese-language environments and is distributed through fraudulent mobile apps. Once installed, it uses Android accessibility-service privileges to take deep control of the infected device.

According to zLabs, RatHat displays fake HTML pages over financial apps to trick users into revealing account credentials. It can also intercept SMS messages and notifications and forward them to attackers, log text changes, extract URLs from the browser address bar, and collect lock-screen PINs, passwords, and unlock patterns.

The most notable feature of RatHat is its AI recognition mechanism. The malware converts real-time device screens into XML layout data and sends it to an unnamed popular AI assistant. The AI model then identifies the center coordinates of UI elements requested by the attacker, determines the text displayed on those elements, and returns navigation instructions such as SCROLL_DOWN, helping attackers quickly analyze and operate infected devices.

Researchers also found that RatHat blocks uninstall attempts: when a user opens the uninstall confirmation screen, the malware intercepts the action and shows a fake uninstall page with misleading error messages. In addition, it includes multiple anti-analysis mechanisms, such as tampering with the APK container, using an unusually large 61MB Android Manifest file, and inserting invalid DEX pseudo-instructions to hinder reverse engineering by security researchers.