Skip to content
AgentProv

2026-09-28 · 2026-Q3

Malicious ChatGPT Custom GPTs on chatgpt.com Push ClickFix and a RAT

Outside the model: Offensive useAgent tooling was not the attack surface, so no layer applies.

Fixable
Architectural
Vendor response
Patched
In the wild
Yes
Actor
Not applicable

Target

Windows users searching for ChatGPT

Impact

Huntress reported on September 28, 2026 that paid Google search ads led victims to a Custom GPT named "Plus 5.6" hosted on the legitimate chatgpt.com domain. The GPT answered every prompt with a "Service Availability Notice" directing users to a "backup domain" on Google Sites, which showed a fake Cloudflare CAPTCHA and a ClickFix PowerShell command. That installed an MSI that sideloaded a DLL through a Canon-signed binary (a second version used Stardock's `DeElevate64.exe`) and delivered a RAT with remote desktop, camera and microphone capture, file search, and payload execution, persisting as "Canon Configuration Reader." Huntress investigated at least 40 incidents tied to the Google Sites page, two confirmed to start at the Custom GPT. OpenAI removed the first GPT on September 25; a replacement with the same name was found on September 27

Root cause

User-published GPTs inherit the AI vendor's domain and reputation, and paid search placement put that trust in front of victims before any content check

Sources

Catalog

Cite this: AgentProv incident record: Malicious ChatGPT Custom GPTs on chatgpt.com Push ClickFix and a RAT (2026-09-28). https://agentprov.com/incidents/2026-09-28-malicious-chatgpt-custom-gpts-on-chatgpt-com-push-clickfix-and-a-rat