Malicious Custom GPT Ads Lure Users into Installing Malware via PowerShell Commands
What Happened — A new scam uses sponsored Google ads that appear to lead to the official ChatGPT site. The ad redirects users to a custom GPT that always returns a “limited availability” notice with a link. Clicking the link opens a fake Cloudflare verification page that prompts the user to paste a PowerShell command, which installs malware on the machine.
Why It Matters for Trust & Control Assurance
- Demonstrates how attackers exploit trusted brand domains to bypass user vigilance, a scenario continuous security‑awareness programs are built to detect and document.
- Highlights the need for verifiable evidence that employees follow safe‑browsing policies and that phishing‑simulation results are tracked over time.
- Aligns with the Security Awareness Training capability, which provides ongoing training, simulated phishing exercises, and audit‑ready logs of user responses.
Who Is Affected — Users of generative AI platforms (enterprise and consumer), SaaS providers integrating AI, and any organization whose staff regularly accesses cloud‑based AI tools.
Recommended Actions
- Reinforce safe‑browsing policies: require direct navigation to
chatgpt.comrather than search‑engine clicks. - Deploy phishing‑simulation campaigns that mimic this custom‑GPT scenario to test and improve user detection.
- Capture and retain logs of suspicious link clicks and PowerShell command executions for audit readiness.
Technical Notes — The scam leverages a sponsored Google ad, a custom GPT hosted on a Google Sites domain, a fake Cloudflare challenge, and a PowerShell payload that drops typical Windows malware. No CVE is involved; the attack vector is social engineering via malicious advertising.