[REVISION HISTORY]

ChatGPT-themed phishing and malware campaigns

Updated 2 times since CLSTR started tracking revisions of this situation.

What changed

2026-10-06 21:20 UTC → 2026-10-08 09:07 UTC · added ~~removed~~

Security researchers have identified evolving phishing campaigns targeting ChatGPT users to steal credentials and deploy malware. Initial reports detailed a campaign where attackers impersonated OpenAI through polished emails. These messages used the official ChatGPT logo and claimed subscription payments had been declined, creating urgency by giving recipients a 48-hour window to ‘Update Payment Information’. This method aimed to capture account credentials and sensitive financial data via fraudulent login pages. Subsequent findings revealed a more sophisticated approach involving fake Google Search advertisements. These ads direct users to deceptive domains that mimic legitimate sites. Once on these pages, users encounter a fraudulent bot, ‘fake ChatGPT Plus 5.6’, which directs them to an alternative site for human verification. This process involves a malicious script that copies code to the user’s clipboard. If the user executes the code via PowerShell or Run commands, a NetSupport Remote Access Trojan (RAT) is installed, granting attackers control over the victim’s computer. More recently, researchers uncovered a human-operated campaign expanding its scope to include other AI tools such as Google Gemini, Claude, Perplexity, and Meta Muse. This operation specifically targets ad account managers and media buyers by promising services like campaign optimization or spend audits. The attackers utilize a ‘browser-in-the-browser’ (BitB) technique, where a fake browser window is drawn inside a legitimate one to spoof trusted domains like accounts.google.com. This allows attackers to trick victims into providing credentials and multi-factor authentication (MFA) codes, enabling them to spend account balances on fraudulent ads or resell access. In this human-managed operation, attackers monitor login attempts and fingerprint devices in real-time. They can manually select which MFA challenge to present to the user, facilitating the hijacking of business accounts and access to payment methods.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.