Skip to content
AI.info

The Pulse

Zimperium reports AI-assisted malware that can restore itself

Zimperium says Android malware called RatHat uses generative AI to navigate device interfaces and can reinstall itself after removal. Its annual research review also reports 608 websites loading Coruna-family code through October 2026.

Zimperium reports AI-assisted malware that can restore itself

Android malware called RatHat uses generative AI to navigate device interfaces and can reinstall itself after removal, according to security company Zimperium. The finding is part of a review of mobile threats targeting credentials, payments and access to devices.

Source prnewswire.com

The company released its 2026 zLabs Year in Review on November 17. It draws on eight selected investigations published between January 1 and November 1, 2026, and examines their implications for enterprise security in 2027.

AI-assisted control

RatHat is primarily distributed through deceptive download sites promoted via malicious advertising, text-message phishing and third-party forums, Zimperium’s earlier technical analysis says. The sites lure people into manually downloading malicious Android apps that appear legitimate.

Its AI component sends a representation of the device’s live accessibility interface to a generative AI assistant. The assistant is used to identify screen coordinates for simulated clicks, retrieve text and signal navigation commands such as scrolling.

According to the researchers, the malware also uses fake interfaces that mimic banking and payment apps to seek financial login credentials, while intercepting authentication codes. It can monitor touch input to reconstruct PINs, passwords and unlock patterns.

RatHat uses Android accessibility features and local Android Debug Bridge pairing to obtain shell-level privileges, the analysis found. A hidden background service can remain after the main app is removed and reinstall it, restoring its malicious permissions.

Web-based iOS attacks

The review also covers a separate route into vulnerable iOS devices: exploit chains delivered through webpages without an app installation. Researchers identified 608 websites loading Coruna-family code through October 2026.

In its separate analysis of the activity, zLabs reported that recovered configurations across 93 of those sites mapped to 21 distinct tracking collectors. The researchers found the code across dating portals, AI-companion platforms, cryptocurrency promotions and streaming services, as well as some compromised legitimate sites using hidden frames.

Recovered infrastructure used different exploit chains depending on the device’s operating-system version, the analysis says. Devices running iOS 17.2.1 or earlier received the older Coruna framework, while later iOS builds received a DarkSword-derived chain.

Payments and corporate accounts

The review also describes threats aimed at financial transactions and enterprise credentials. PixRevolution can redirect PIX payments, while RecruitTrap uses recruitment-themed phishing to seek corporate credentials, according to Zimperium.

Other cases combine surveillance with remote control or extortion. Mantax Otax combines spyware and ransomware functionality targeting older Android versions; Arsink RAT uses widely available cloud services to support surveillance and data theft.

The review also points to risks from AI capabilities embedded in mobile applications and from AI-generated code. Those issues were highlighted in Zimperium’s 2026 Global Mobile Threat Report, which calls for security oversight of AI-generated code.

Sources

  1. PR Newswire
  2. Zimperium
  3. Zimperium

Related stories

Explore

More articles