ClickFix Style Attack Uses Grok, ChatGPT for Malware Delivery

A new ClickFix-style attack blends SEO poisoning with legitimate AI domains to deliver infostealer malware, exploiting trusted platforms to mislead users into running malicious prompts. Huntress researchers Stuart Ashenbrenner and Jonathan Semon describe a technique where a user searching for a help problem is shown first-page results to legitimate ChatGPT and Grok links, then guided to a conversation that appears trustworthy but instructs the user to run a command that installs the AMOS MacOS stealer.

The attack was observed affecting a Huntress customer on 5 December 2025, with the malware harvesting data such as credentials and browser data while maintaining persistence. According to Huntress's blog post, the attacker engineers the LLM prompt to mimic legitimate troubleshooting and creates a shareable URL from the AI platform, then propagates the malicious conversation across content farms and forums to inflate legitimacy.

The campaign relies on social trust in AI chatbots and search engines, rather than on traditional download prompts or warnings. Defenders are advised to look for behavioural anomalies such as osascript requests and hidden executables in user home directories, and users are urged to avoid executing terminal commands from unfamiliar sources.