Cybercriminals have found a new way to exploit the popularity of generative AI tools by deceiving users into downloading malware. A recent case highlights a fake Google Gemini installer being used to distribute the Vidar information stealer, compromising browser credentials and other sensitive data.
How the Attack Unfolded
Unlike traditional phishing attacks, this incident utilized a seemingly legitimate software search and download method. This strategy demonstrates how routine searches for AI tools can be manipulated to facilitate credential theft. This mirrors other campaigns where fake AI-themed downloads have been used to distribute similar malware.
Darktrace analysts detected this activity in July 2026, within a client environment across the Europe, Middle East, and Africa region. Their report, shared with Cyber Security News, linked the suspicious executable to Vidar after observing unusual behaviors and suspicious outbound connections.
Exploiting Trusted Platforms
The infection chain began when a user executed a file named Download_Google_Gemini_For_Windows.exe from their Downloads folder. A search for this filename led to a Google Colab page, which made the campaign appear legitimate. Google Colab, favored by developers and researchers, served as a credible staging ground for the attack.
The page redirected users to a secondary site masquerading as a “Windows Software Hub,” where the fake installer was available for download. Although the source of the initial download could not be confirmed, encrypted sessions with Google Colab suggested user interaction before reaching the secondary site.
Implications for Security
Once executed, the program was identified as a new Go-compiled variant of Vidar, which communicated with suspicious external systems. The malware aimed to collect valuable data from infected devices, including saved passwords and browser session information, potentially granting attackers unauthorized access to various accounts.
Darktrace’s swift response involved blocking communications with the malicious infrastructure and isolating the compromised endpoint. This action was based on abnormal behavior rather than the perceived legitimacy of the installer or hosting platform.
Lessons and Recommendations
This incident underscores the need for caution when downloading AI-related software. Users are advised to obtain such tools only from verified sources and to avoid downloads that navigate through unfamiliar pages. Additionally, instructions to disable antivirus protection should always raise red flags.
Security teams must monitor for unusual browser credential access, newly launched executables from Downloads folders, and rare encrypted connections from recent installations. These measures are vital as attackers continue to exploit AI interest with misleading search results and fake installation guides.
Ultimately, this campaign demonstrates how attackers leverage trusted platforms and social engineering to distribute information stealers through seemingly credible channels. The broader threat landscape shows that the demand for AI tools can easily be manipulated as a lure for credential theft.
