Cybercriminals are deploying counterfeit installers masquerading as genuine artificial intelligence applications, including OpenAI ChatGPT and InVideo AI, to spread ransomware and destructive malware.
Cisco Talos researchers uncovered that CyberLock, a dangerous ransomware built with PowerShell, primarily seeks out and encrypts specific files, while another variant, Lucky_Gh0$t, is a modified form in a long line of malware derived from Yashma and Chaos families.
The campaign’s latest addition, the Numero malware, disables computers by tampering with the graphical interface elements in the Windows operating system.
Those behind these schemes have set their sights on professionals and companies active in marketing and business sales, taking advantage of the popularity of legitimate AI platforms in these sectors.
One of the schemes involves a fraudulent website named to imitate a well-known lead generation service, NovaLeads, which draws in visitors using trickery meant to boost its appearance in search results.
Visitors enticed by offers of free trials and paid subscriptions instead download a packaged program that launches CyberLock ransomware, which then tries to gain admin status before encrypting files across multiple drives, demanding a ransom of $50000 in Monero within three days.
Malicious Campaigns Hide Behind AI Brands
Cybersleuths noted that the ransom note justifies the demand by claiming proceeds will aid people in regions like Palestine, Ukraine, and several others facing hardship, insisting the attack is necessary to prompt action for humanitarian causes.
The attackers take additional steps by removing remnants of deleted files to complicate any forensic recovery and erase evidence from disks using built-in Windows utilities.
Another tactic observed was packaging Lucky_Gh0$t ransomware within a mock premium ChatGPT installer, even including authentic AI tools to appear credible while launching a harmful executable that imitates Microsoft’s system files.
If activated, this installer launches a ransomware attack that targets smaller files for encryption after wiping local backups, then presents the victim with a ransom note and a unique identifier, directing them to communicate payment details over a private messaging platform.
Threat actors have also exploited interest in the InVideo AI platform to distribute Numero, a disruptive program that repeatedly executes in a cycle, checking for security tools and repeatedly corrupting the desktop interface of Windows PCs, making systems unusable.
Compiled in January, Numero cycles through a process that rewrites the Windows interface with rows of numbers, rendering machines effectively bricked.
Researchers revealed parallel criminal activity involving fraudulent artificial intelligence tool installers promoted through ads on major social sites: users are led to imitation video generation tools and prompted to start video projects, only to be infected with multi-stage malware.
Victims unwittingly download a payload known as STARKVEIL, which acts as a launchpad for additional information stealing programs—capable of harvesting credentials, surveying machines for crypto wallets, and remotely controlling infected systems.
Recent analyses have linked these sophisticated tactics to a group believed to operate from Vietnam, relying on multiple strains of malware to evade detection and broaden their reach.
Security experts caution that such schemes extend beyond the creative industry, with anyone tempted to try new AI tools potentially at risk.
