Cybercriminals Manipulate Search Engines, Leveraging ChatGPT and Luma AI Trends to Distribute Malicious Payloads

In a striking escalation of cybercrime tactics, threat actors are now exploiting the surging popularity of artificial intelligence tools, such as ChatGPT and Luma AI, to weaponize search engine results and deliver advanced malware to unsuspecting victims.

A recent investigation by Zscaler’s ThreatLabz has revealed a sophisticated campaign in which cybercriminals leverage Black Hat SEO techniques to manipulate search rankings, ensuring their malicious sites appear prominently for trending AI-related queries.

Black Hat SEO and Multi-Layered Redirection

The attackers’ strategy begins by creating AI-themed websites, often built on common platforms like WordPress, and optimizing them for high visibility in search results.

When users search for terms like “Luma AI blog,” these fraudulent sites frequently appear among the top results.

Once a user clicks through, they land on a seemingly legitimate webpage, but lurking beneath the surface is a complex redirection and malware delivery mechanism.

JavaScript embedded in these sites initiates a multi-step process: it collects browser data such as version, window resolution, cookies, and user agent, then encrypts this information using a randomly generated XOR key.

The data is transmitted to an attacker-controlled domain, typically disguised as a legitimate service.

The server validates the information and responds with a 302 redirect to an intermediate site, which further checks the user’s public IP and other parameters before ultimately directing them to a malware payload.

Notably, these sites employ sophisticated detection techniques to check for the presence of ad blockers or DNS guards that could disrupt their redirection chain.

If detected, the script aborts the attack, making the campaign more resilient to automated defenses.

The configuration details for these redirections are often stored in Base64-encoded strings with a custom character set, further obscuring the malicious domains and complicating detection efforts.

Malware Delivery and Evasion Tactics

The final payloads in these campaigns include notorious malware families such as Vidar Stealer, Lumma Stealer, and Legion Loader.

These are often delivered as password-protected ZIP archives, with the password provided on the final download page.

The archives contain large NSIS installer files, sometimes exceeding 800MB, to bypass sandbox detection systems that have file size limitations.

Upon execution, the NSIS installer deploys files with a .docm extension, masquerading as Microsoft Word documents but serving as components of the malware.

These are combined to generate an AutoIT loader executable, which in turn delivers the final payload.

To evade antivirus detection, the malware checks for and terminates processes associated with popular security software, including Quick Heal, Webroot, Sophos, BitDefender, Avast, AVG, Norton, and ESET.

Legion Loader employs a multi-stage attack chain, utilizing an MSI file that installs itself in the AppData directory and deploys decoy software.

It collects and transmits system information to a command-and-control (C2) server, retrieves a dynamic password for payload extraction, and utilizes DLL sideloading and process hollowing to inject malicious code into legitimate processes, such as explorer.exe.

This technique allows the malware to execute shellcode that deploys cryptocurrency-stealing browser extensions.

A Call for Vigilance

The campaign’s indicators of compromise include domains like chat-gpt-5.ai, luma-ai.com, and krea-ai.com, as well as a range of C2 servers and file hashes associated with the malware payloads.

Security researchers have mapped these tactics to several MITRE ATT&CK techniques, including drive-by compromise, DLL sideloading, process injection, and exfiltration over C2 channels.

As the popularity of AI tools continues to grow, cybercriminals are increasingly exploiting these trends to lure victims.

Users are advised to exercise caution when clicking on search results related to AI topics, verify the legitimacy of websites, and ensure their security software is up to date.

The threat landscape is evolving rapidly, and vigilance is essential to stay one step ahead of these sophisticated attacks.