A newly discovered Android malware called RatHat is using artificial intelligence to help attackers navigate infected phones, steal banking credentials and maintain remote control even after victims try to remove it.
Security researchers at Zimperium’s zLabs said the malware combines several established Android attack techniques with an AI-powered automation system. That allows it to react to what appears on a victim’s screen instead of relying entirely on fixed instructions.
RatHat is being spread mainly through malicious advertisements, SMS phishing messages and fake websites that encourage users to download Android APK files from outside Google Play.
Researchers believe the operators may be linked to China after finding Chinese-language prompts inside the malware’s AI workflow. That attribution remains an assessment based on technical evidence, not a confirmed identification of the people behind the operation.
Proposal for Conducting Cyber Crisis Drill, Tabletop Exercise (TTEx) & CCMP Readiness Exercise
How RatHat takes control of an Android phone
The attack begins with social engineering. A victim is pushed towards a fake download page and persuaded to install an app manually.
Once installed, RatHat asks for Android Accessibility permissions. Accessibility features are designed to help people interact with their phones, but malicious apps can abuse them to read what appears on the screen and simulate taps, swipes and other actions.
RatHat then goes further. Researchers found that it can enable Android’s Developer Options and Wireless Debugging features before pairing with the phone through Android Debug Bridge, commonly known as ADB.
ADB is a legitimate developer tool normally used to communicate with Android devices during testing. In RatHat’s case, the feature is abused to gain a more powerful shell-level environment inside the phone.
The malware installs additional components that can execute commands, bypass some battery restrictions and help keep the infection alive. Researchers say these components can restore each other if either the main malware or one of its supporting agents is removed.
That makes a simple uninstall attempt potentially ineffective.
The AI layer changes how the malware navigates
Traditional mobile malware often follows pre-programmed instructions. If a banking app changes where a button appears, those instructions may stop working.
RatHat uses a different approach.
Zimperium found that the malware can convert information about the phone’s visible interface into structured data and send it to an AI assistant. The system can then identify buttons or text, determine where an item is located and issue instructions such as scrolling through the screen.
In simple terms, instead of being told “tap the button at this exact spot”, the malware can ask an AI system to work out where the relevant button is.
That could make the automation more adaptable across different devices, applications and screen layouts.
This does not mean the AI independently “hacks” the phone. RatHat still depends heavily on traditional techniques such as phishing, sideloading, Accessibility abuse and ADB exploitation. AI mainly makes the remote-control process more flexible.
Banking passwords, OTPs and lock-screen codes at risk
RatHat can display fake login screens over legitimate banking and cryptocurrency applications. These overlays are designed to trick victims into entering usernames, passwords and other financial credentials.
Researchers also found capabilities for intercepting SMS messages and notifications, which can expose one-time passwords and two-factor authentication codes.
The malware can monitor text entered on the phone and collect information from browser address bars.
According to Malwarebytes’ review of the research, RatHat can also capture touch coordinates and use them to reconstruct certain PINs or unlock patterns.
It can further interfere with removal attempts. Zimperium found that RatHat may detect when a user reaches the uninstall confirmation screen, cancel the process and show a fake Google Play error message instead.
AI-powered Android threats are no longer theoretical
RatHat is part of a wider shift towards AI-assisted mobile malware.
Earlier this year, researchers examined SurxRAT, another Android remote-access Trojan that could use large-language-model components to automate malicious activity and interact with device interfaces.
Android malware has also increasingly abused sideloaded applications, fake download pages and elevated permissions to target banking users. Zimperium’s mobile threat research has repeatedly identified sideloaded apps and mobile phishing as major Android attack routes.
RatHat shows how these familiar methods can become more dangerous when combined with adaptive AI.
The strongest defence still begins before installation. Avoid APK files received through SMS, advertisements, Telegram groups or unfamiliar websites, and be especially suspicious when an ordinary app requests Accessibility access or asks you to enable Developer Options or Wireless Debugging.
What this means for you: If an app you downloaded outside Google Play asks for Accessibility access or Developer Options, stop immediately. If you suspect RatHat-level compromise, backing up essential data and performing a factory reset may be safer than relying on a normal uninstall.
Follow for daily updates on cybercrime, corporate fraud, DFIR, hacking, investigations, and digital forensics
