I came across an interesting Huntress investigation and thought it was worth writing up briefly for anyone who works with endpoints or supports users day to day.
Apparently, attackers have been using malicious ChatGPT Custom GPTs as part of an attack involving ClickFix and a remote access Trojan (RAT).
How the chain worked
What caught my attention was how the whole thing fitted together.
The attackers created a Custom GPT disguised as a ChatGPT upgrade called “Plus 5.6”. Users were then directed to another website, supposedly because the main service was unavailable.
That website displayed a fake Cloudflare verification page and instructed users to copy and run a command on their computers.
And that is where things get nasty.
This is a technique known as ClickFix, where attackers trick users into running malicious commands themselves rather than simply exploiting a vulnerability. In this case, the commands triggered PowerShell activity that eventually led to the installation of a RAT.
Depending on its capabilities, this type of malware can allow attackers to access files, monitor activity, and even use the victim’s camera or microphone.
I am not claiming to have dissected the samples myself. Huntress published a detailed write-up of the campaign, including how a legitimate-looking ChatGPT experience funnels people into that ClickFix page. If you want the full technical chain, their post is the place to start.
What stood out to me
A few things stood out from this case:
- Familiar platforms are not always a guarantee of safety. Attackers can abuse users’ trust in legitimate services. A page that looks like it belongs on chatgpt.com can still be an attacker-built Custom GPT.
- Be careful about commands you copy and run. A website asking you to paste something into PowerShell or the Windows Run dialogue should raise questions, especially if you do not understand what the command does.
- A valid digital signature does not automatically mean something is safe. It is important to look at what a programme actually does, not just whether it is signed. Signed binaries can still be abused as part of a longer infection chain.
Takeaway
Working in IT, I find cases like this a useful reminder that cybersecurity is not only about having the right tools in place. Understanding how attacks work, recognising suspicious behaviour, and helping users avoid common traps matter just as much.
If a user ever says a website asked them to “verify” by running a command, treat that as a red flag and investigate before anyone pastes anything.
Definitely worth a read if you are interested in cybersecurity: Attackers Abuse ChatGPT Custom GPTs to Deliver RAT via ClickFix (Huntress).
Originally shared on LinkedIn.