OpenAI patches ChatGPT Lockdown Mode after researchers bypass its safeguards
Reported by The Hacker News, CyberSecurityNews and TechRadar3 sources ↓
Check Point Research found that HTML previews could open popups outside Lockdown Mode’s sandbox, potentially exposing sensitive data. OpenAI added server-side safeguards to block the workaround.
Key points
- The flaw could let attackers bypass Lockdown Mode through popups opened from HTML previews
- Researchers said the bypass could enable data theft, phishing or malware delivery
- OpenAI patched the issue on September 18, 2026, according to The Hacker News
Sources
- The Hacker NewsRead ↗Researchers Bypass ChatGPT Lockdown Mode to Exfiltrate Chats Without User Interaction
- CyberSecurityNewsRead ↗Hackers Could Bypass ChatGPT Lockdown Mode to Steal Conversations Without Users Knowing
- TechRadarRead ↗'AI security controls are only as strong as the wider application environment around them': OpenAI forced to patch after experts bypass ChatGPT 'lockdown mode' to gain access to entire chat history
Details
Check Point Research found that nested frames in ChatGPT’s HTML previews could access popup functions that escaped the preview’s sandbox. The popups were no longer subject to the same restrictions, including controls on access to data and external connections.
The researchers said an attacker could exploit the flaw through a crafted prompt, malicious GPT or shared chat. The popup could be used to expose chat data, present phishing content or deliver malware; The Hacker News reports that the flaw also allowed potential zero-click data exfiltration in the desktop app.
OpenAI added safeguards to block the workaround. The Hacker News reports that the fix was applied on September 18, 2026, while TechRadar describes it as a server-side patch.
Written from coverage as of
Images
- The Hacker News
- TechRadar
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh