Malicious Custom GPTs Used as RAT Delivery Lure in ClickFix-Style Campaign

A new campaign identified by security analysts leverages malicious Custom GPTs to deliver malware to unsuspecting users. According to reporting, threat actors have created and promoted Custom GPTs on the OpenAI platform that serve as lures, directing victims to Google domains under the pretense of AI-related functionality. The attack flow follows a ClickFix-style pattern: users are persuaded to execute commands or follow instructions that, rather than performing an AI task, execute payloads resulting in remote access trojan (RAT) infection. The use of legitimate OpenAI and Google domains helps the social engineering effort, as victims are more likely to trust instructions originating from recognized AI and search properties. As of the report date, no specific CVE, zero-day exploit, or software vulnerability has been identified. The threat relies on social engineering and the trust users place in AI chat interfaces. OpenAI and Google are reportedly monitoring the platform and removing malicious GPTs, but the dynamic nature of Custom GPT creation makes comprehensive detection challenging. Security experts recommend that users avoid executing unknown commands or scripts from AI chat interfaces and that organizations educate users about the risks of interacting with unfamiliar Custom GPTs.