RESEARCHERS have observed a new campaign that distributes the AsyncRAT remote access trojan through counterfeit artificial intelligence learning guides, luring victims who seek free tutorials on topics such as deep learning, neural networks and data science. The attackers bundle the malware inside seemingly legitimate 7z archives that claim to contain lecture notes, code samples and video links, thereby exploiting the current hype around generative AI to increase the likelihood of a successful download. Once executed, AsyncRAT gives attackers full remote control over the infected machine, enabling them to harvest credentials, capture screenshots and exfiltrate sensitive files.
The infection typically begins with a seemingly benign 7z archive that contains a Windows shortcut file (.lnk) and a pair of PDF documents that appear harmless but are used to hide malicious scripts. When the shortcut is double‑clicked it executes a series of obfuscated PowerShell commands that download a second‑stage payload from a remote server, which then launches AutoHotkey to masquerade as a legitimate automation tool. This second stage writes a scheduled task registered under a generic service name, ensuring the trojan runs at each logon and survives reboots while maintaining low visibility to conventional antivirus scanners.
The PowerShell payload establishes contact with a command and control server over HTTP or HTTPS, sending basic system information and awaiting further instructions. AsyncRAT’s modular design lets operators load plugins for keylogging, webcam access, file system browsing and credential theft from browsers and email clients. The malware also injects code into trusted processes such as explorer.exe to avoid heuristic detection. All communications are encrypted and often routed through compromised legitimate domains to blend with normal traffic, making detection by network‑based intrusion systems more challenging.
Although no CVE identifier has been assigned to this activity, FortiGuard Labs notes that the campaign has been active from early June 2026 and primarily targets Windows users who download the fake guides from file‑sharing forums or social media posts promising exclusive AI course material. Telemetry from multiple victim environments shows consistent beaconing to domains hosted on bulletproof hosting providers.
The lure files are often named with terms like “AI_Machine_Learning_Course.7z” or “DeepLearning_Tutorial_Pack.7z” to appear credible, and the attack relies on social engineering rather than software vulnerabilities, which explains the absence of a CVE.
Defenders should block the execution of unsigned PowerShell and AutoHotkey scripts through application control policies, and enforce script logging to capture any attempt to run obfuscated code. Monitoring for newly created scheduled tasks that point to unusual file paths or use generic service names can reveal persistence mechanisms, while restricting the extraction of archives from untrusted sources reduces the initial infection vector.
Additionally, deploying endpoint detection and response solutions that flag process injection into trusted binaries will help identify the stealth techniques used by AsyncRAT. Educating users to verify the provenance of AI learning materials and to avoid opening archives from unknown senders further reduces risk.
Organizations should also maintain offline backups of critical data, segment networks to limit lateral movement if a host becomes compromised, and subscribe to threat intelligence feeds that provide indicators of compromise for AsyncRAT, such as the specific SHA‑256 hashes of the malicious 7z archives and the domains used for command and control. Keeping security tools up to date and conducting regular purple‑team exercises will ensure defenses stay effective against this evolving threat.