This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A malware campaign is using fake AI-themed websites, impersonating legitimate tools like Luma Dream Machine and CapCut, to distribute a new infostealer called 'Noodlophile'. Attackers lure users via social media to upload images for 'AI processing', then provide a malicious ZIP file instead of the promised media. The malware, which is sometimes bundled with the XWorm remote access trojan, steals browser credentials, cookies, and cryptocurrency wallet data, exfiltrating it via Telegram bots.
A campaign reportedly used fake AI video generation sites to distribute malware under the guise of AI-generated content. Promoted via social media, these sites allegedly tricked users into downloading files containing Noodlophile Stealer, a previously unreported infostealer, and in some cases XWorm. The malware harvested credentials and could enable remote access.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
Also classified as
The Noodlophile malware deployed via the scam is specifically designed to steal sensitive personal data, including browser credentials, cookies, and crypto wallets.
The incident involves threat actors setting up fake AI platforms and social media pages to scam users into downloading malware under the guise of legitimate AI services.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The incident is a social engineering campaign orchestrated by human threat actors who intentionally exploit the popularity of deployed AI systems to distribute malware.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
This incident is a financially motivated cybercrime campaign leveraging the popularity of generative AI tools as social engineering lures to distribute information-stealing malware. While it poses a threat to individual privacy and digital security, it has minor national security implications and can be managed through standard cybersecurity practices.
The reports describe users seeking 'AI tools for video and image editing' and 'text-to-video AI service'. These systems operate as tools that perform specific, deterministic functions on direct input without autonomous planning or agency.
Minimal or No Risk: The AI systems being impersonated (Luma Dream Machine and CapCut) are general-purpose image and video generation tools used for entertainment and content creation, which pose minimal or no risk under the EU AI Act.
The impersonated AI systems, such as Luma Dream Machine, are designed for generating and editing videos and images from user prompts.
The reports from multiple cybersecurity sources provide highly detailed and consistent technical analyses of the malware campaign, its delivery mechanisms, and the specific AI tools being impersonated. There is little ambiguity regarding the facts of the campaign.
None. The campaign is clearly a malicious cyberattack leveraging AI hype as a social engineering lure.
Technology
Deceptive Executable Extensions
Exploited Windows disabled file extensions to hide exe as mp4.
Evasion of Static Scanners
Large binary size and code obfuscation bypassed traditional scanners.
Memory-Based Payload Execution
Executing malware in-memory to evade traditional disk-based detection.
Data Inputs
Malicious Prompt Upload Portals
Fake AI sites accepted user files to mimic real AI processing.
Fake Video Generation Outputs
Delivering malicious ZIP files disguised as AI-generated video outputs.
Obfuscated Script Inputs
Using division-by-zero lines to break automated AST parsers.
Human Factors
Exploitation of AI Hype
High public interest in generative AI made users trust fake platforms.
Disabled File Extensions
Windows users unable to see the true exe extension of the file.
Trust in Social Media Verification
Verified badges on fake Facebook pages increased perceived legitimacy.
Process and Methods
Lack of Platform Moderation
Social media platforms failed to quickly remove fake AI advertisements.
Lack of Download Verification
Users downloading files from unverified third-party websites.
Inadequate Security Training
Small businesses and freelancers lacked training to spot fake AI lures.
Management
Lack of Strict Activity Separation
Businesses failed to separate business tasks from personal AI testing.
Regulatory Environment
Weak Cybercrime Enforcement
Thriving regional cybercrime ecosystems operate with minimal enforcement.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
