How Are Hackers Using Custom ChatGPT Bots in Attacks?
Attackers are increasingly leveraging the trusted ChatGPT platform by creating fake custom bots that appear legitimate due to their hosting on the official chatgpt.com domain. These custom bots, designed to serve specific purposes or personas, are exploited to deliver a malicious lure. For example, a fraudulent bot named "Plus 5.6" mimics OpenAI's model naming conventions to avoid suspicion and immediately directs users to a secondary, malicious site.
What Is the Role of Google Sites in This Scam?
The attackers host a deceptive "backup domain" on Google Sites, a legitimate web creation service, to enhance the credibility of the scam. This site presents users with a fake Cloudflare CAPTCHA, prompting them to execute code on their Windows PC under the guise of fixing an issue. This social engineering technique, known as a ClickFix attack, convinces victims to perform actions that enable malware installation.
What Kind of Malware is Deployed and What Are Its Capabilities?
The downloaded malware is a Remote Access Trojan (RAT) called '@input,' which grants attackers extensive control over the infected computer. Once installed, it allows surveillance through the webcam and microphone, the ability to view and manipulate files, monitor running programs, and assess network connections. The RAT can also stealthily download additional malicious components to further compromise the system, communicating with its operators through encrypted channels disguised as regular internet traffic.
Who Is at Risk and How Widespread Is This Threat?
Primarily Windows users engaging with unexpected ChatGPT links are at risk, especially those who may not scrutinize URLs carefully. The campaign has already compromised dozens of victims, with security teams responding to multiple incidents centered around the same malicious Google Sites domain. The rapid reappearance of new malicious custom GPTs after takedown efforts highlights the persistence and adaptability of the attackers.
What Are the Key Takeaways for Users and Security Professionals?
Users should remain vigilant when interacting with links, even if hosted on reputable domains like chatgpt.com or platforms such as Google Sites. Any prompt requesting to run code or commands outside the expected application environment should be treated with suspicion. Security professionals must consider that attackers are exploiting trusted AI service features and mainstream hosting platforms to enhance phishing campaigns. Continuous monitoring for unusual activity and educating users about these evolving tactics are essential to reduce infection risk.