A new malvertising campaign is exploiting ChatGPT’s popularity by promoting a weaponized fake download site through sponsored search results, delivering malware to both Windows and macOS users.
Researchers from Evalian’s SOC team uncovered the operation, which uses convincing OpenAI branding and paid search ads to attract users who are actively looking for legitimate AI tools.
The campaign centers on a malicious domain, openew[.], that hosts a site crafted to closely resemble an official ChatGPT download page.
Visitors are offered several download options, including installers for Windows and macOS as well as a Chrome extension.
While the browser extension simply redirects users to a legitimate listing to establish trust, the Windows and macOS installers contain trojanized payloads.
The domain was recently registered through Namecheap and resolves to the IP address 144[.]172[.]104[.]205, which is hosted on RouterHosting infrastructure — a provider frequently observed in short-lived malicious campaigns.
The Windows payload is distributed as Chat_GPT.exe (SHA256: 56CC26E88C064B0C423AA8AD6530E58F91D1E4D28FAB1A8BCEDEF16A6582B4D2). It uses an Inno Setup installer to deploy an Electron-based application.
Although the binary looks legitimate at first glance, static analysis reveals inconsistencies such as mismatched metadata and a code-signing certificate issued to an unrelated organization, F.F.A.P. Hurkmans Beheer B.V. This demonstrates how valid-looking signatures can be abused to lower user suspicion without guaranteeing software authenticity.
Fake ChatGPT Site Spreads Malware via Ads
Static inspection shows the application bundles a Chromium-based runtime with an obfuscated JavaScript payload stored in the app.asar archive. A large script, identified as winter.js, contains heavily obfuscated logic that relies on encoded strings and dynamic execution patterns, which complicates straightforward analysis.
The application includes Node.js modules such as child_process, fs, and systeminformation, indicating capabilities for system reconnaissance, file operations, and executing commands.
Dynamic analysis shows the malware uses CAPTCHA-based gating before it runs its main routines, a technique aimed at evading automated sandbox and bot detection. After a victim completes the CAPTCHA, the malware spawns multiple PowerShell processes with execution flags such as -ExecutionPolicy Unrestricted, which suggests staged payload delivery where commands are injected at runtime rather than being statically embedded.
Evalian’s SOC team observed that the malware creates a Chromium-style profile in %AppData%Satoshi to maintain persistence and store artifacts like cookies and cache files. This, together with event-driven execution, indicates the malware delays its primary actions until specific user interactions occur, making detection more difficult.
The embedded network configuration references legitimate DNS-over-HTTPS providers such as Cloudflare and Google, allowing malicious traffic to blend into normal encrypted DNS flows and helping to obscure command-and-control communications from traditional network monitoring tools.
The macOS variant (SHA256: 7E5B708F6659B1FAD3AAE7B589A706434FBF21708AEEC5AF5910189B96E25FEF) was mostly undetected by antivirus engines at the time of discovery, which may indicate either limited distribution or effective evasion methods.
This campaign illustrates how threat actors are advancing malvertising techniques by combining trusted branding, modern application frameworks like Electron, and layered evasion tactics such as obfuscation, CAPTCHA gating, and staged execution. Unlike traditional phishing, malvertising targets users with high intent, increasing the likelihood of initial compromise.
For defenders, key indicators include unexpected Electron applications launching scripting engines, mismatched installer metadata, and unusual directories such as %APPDATA%Satoshi. Monitoring newly registered domains that impersonate software vendors and focusing on process behavior rather than signatures alone are critical defenses.
As AI tools become more widely adopted, campaigns that impersonate well-known brands will likely increase, underscoring the need for stronger user awareness and behavioral detection controls.
Follow us on Google News, LinkedIn, and X to get more instant updates.
