Attackers published a custom GPT within ChatGPT named ‘Plus 5.6,’ designed to look like a new model release, and promoted it via Google ads appearing atop search results for ‘chatgpt.’ The GPT told users the main ChatGPT domain was unavailable and redirected them to a lookalike site that used a fake CAPTCHA to trick victims into pasting a malicious command into Windows Terminal, triggering a multi-stage infection chain that installed remote-access malware. At least 40 victims were affected before OpenAI removed the GPT, though operators relaunched a replacement within two days.