Lockdown Mode Gets Bounced Out the Side Door: ChatGPT's Sandbox Let Chats Leak Silently
Ourtake
The bouncer was posted at the front door while the sandbox left the back window open.
What happened
Check Point Research found a flaw in ChatGPT's HTML preview sandbox that let attackers bypass OpenAI's Lockdown Mode. Attacks via a malicious prompt, custom GPT or shared chat could silently exfiltrate chat history, spoof windows for phishing and malware delivery, and steal credentials. OpenAI removed the permissive popup permissions, and Check Point confirmed the technique no longer works.