threatdeck
Live 14:45 UTC ·

AI Stack Alert High TD-2026-0017 For CISO, SOC / IR, AppSec, AI Eng AI-assisted

ChatGPT Lockdown Mode bypassed via HTML preview popups; OpenAI has since patched

Check Point Research found a way to bypass OpenAI's ChatGPT Lockdown Mode using the HTML page preview feature. The bypass allowed silent chat-history exfiltration, phishing popups, and malware delivery. In the Electron-based ChatGPT Classic app, the popup technique worked without a user gesture. OpenAI says it removed the permissive popup sandbox permissions, and Check Point confirmed the technique no longer works.

Why it matters

Users who enabled Lockdown Mode to protect sensitive data could have had chat history and uploaded files sent to an attacker's server. Attackers could also show spoofed ChatGPT windows that pushed fake updates or credential-stealing re-login pages.

Do this

  1. Confirm that ChatGPT Classic and the browser client are on the current version that includes OpenAI's HTML-preview sandbox fix.
  2. Do not rely on Lockdown Mode alone for highly sensitive data. Limit what users paste or upload into ChatGPT until the patch is verified across all clients.
  3. Instruct users not to click links or download 'updates' that appear inside ChatGPT previews or shared chats, and to verify any update through the official vendor site.
  4. Review SOC or proxy logs for unexpected outbound requests from ChatGPT Classic processes or unusual popup-driven sign-outs and re-authentication prompts.
  5. Treat shared ChatGPT chats and custom GPTs from untrusted sources as untrusted content.

Signals

  • NoCISA KEV
  • 36/100Priority
  • 80%Confidence

Evidence

“OpenAI recently introduced Lockdown Mode, an optional advanced security setting designed to protect sensitive user data from prompt injection-based exfiltration attacks.”

Check Point Research

“Check Point Research discovered a vulnerability that allowed attackers to bypass Lockdown Mode, rendering its core protections ineffective.”

Check Point Research

“We found that window.open in the Electron-based ChatGPT Classic application works even without a user gesture.”

Check Point Research

“On September 18, 2026, OpenAI informed us that it had deployed a fix by removing the allow-popups and allow-popups-to-escape-sandbox permissions from the outer HTML-preview sandbox iframe.”

Check Point Research

Sources

Affected / related

ChatGPT, ChatGPT Classic, Lockdown Mode, OpenAI, HTML preview sandbox escape via popups

Published 2026-11-05 14:00 UTC · updated 2026-11-10 00:24 UTC · AI-assisted summary, checked against quoted sources · How we score

Following

Follow the vendors, products, AI tools, techniques and actors you care about. Matching stories, CVEs and exploited bugs are marked, a Following tab appears in the feed, and returning visits show what changed. Saved only in this browser.

Presets

Get updates

A personal RSS feed of new stories on everything you follow. Paste it into any feed reader, Slack or Teams RSS app.