Global Edition
Global Edition
UK Edition
EU Edition
US Edition

Understand the story, not the spin.

Markets

Malicious ChatGPT GPTs led users to a remote access trojan

Published 30 September 2026 · All stories from this day

Malicious ChatGPT Custom GPTs directed users to a fake CAPTCHA and a command-based malware lure that installed a remote access trojan, according to Huntress. The cybersecurity firm said it responded to at least 40 incidents tied to a Google Sites domain used in the campaign, but confirmed that only two began through a malicious Custom GPT. The broader incident count does not represent 40 infections through the GPT feature.

The attackers created a GPT named Plus 5.6 that appeared to offer a legitimate ChatGPT service. Some users reportedly reached it through sponsored Google search results. When opened, the GPT displayed a false notice that the service was limited on the primary domain and directed users to a backup page hosted on Google Sites.

That page imitated a Cloudflare CAPTCHA and prompted visitors to copy and run a command, a social-engineering technique known as ClickFix. Huntress said the command launched PowerShell, which retrieved an obfuscated script and silently installed a malicious Windows installer. Running the command themselves could make users less likely to recognize the activity as an attack.

The first reported delivery chain used signed Canon software to...

Now playing
0:00 / 0:00