Last Updated: 2026-06-18
What’s Happening
Malvertising of AI tooling to direct users to geoshitties (my nickname of free subdomain web hosting) Gitlab[.]io (typically used to host a projects documentation) making it a high-regret block for IT users.
In this article shared Claude chats were being utilized, making it abuse of a Living off a Trusted Site (LOTS).
Actions
- Link to the article’s IOCs to hunt or block in your organization.
https://documents.trendmicro.com/assets/txt/Indicators%20of%20Compromise%20(IOCs)-n9ZfTcK.txt
-
Perform ClickFix mitigations where possible for Windows and Mac.
-
A couple weeks ago I rooted around and dug up the share URL formats to a bunch of AI platforms to hunt/block.
ChatGPT https://chatgpt.com/share/{id}
Claude https://claude.ai/share/{id}
Copilot https://copilot.microsoft.com/shares/{id}
DeepSeek https://chat.deepseek.com/share/{id}
Gemini https://g.co/gemini/share/{id}
Grok https://grok.com/share/{id}
Manus https://manus.im/share/{id}
Meta AI https://www.meta.ai/s/{id}
Poe https://poe.com/s/{id}
Qwen https://chat.qwen.ai/s/{id}
Unfortunately, I also found many AI platforms don’t have a specific share URL format; the chat ID is simply made public.
HuggingChat https://hf.co/chat/r/{id}
Character.AI https://character.ai/chat/{id}
T3 Chat https://t3.chat/chat/{id}
Mistral Le Chat https://chat.mistral.ai/chat/{id}
Lovable https://lovable.dev/projects/{id}
Bolt https://bolt.new/~/{id}
NotebookLM https://notebooklm.google.com/notebook/{id}
Perplexity https://www.perplexity.ai/search/{id}
v0 https://v0.app/chat/{id}
The use of implicitly trusted platforms like search engines, free hosting and LLMs make it difficult to deploy resilient posture and can often encounter high-friction with the business.
Does your org need every AI platform available? And how important is sharing? In cases like these ,proxy and web logs with HTTP referrer data make excellent risk-based detections.
