Researchers Bypass ChatGPT Lockdown Mode to Exfiltrate Chats Without User Interaction
Summary: Check Point researchers disclosed a now-patched ChatGPT vulnerability that bypassed Lockdown Mode, the optional setting OpenAI introduced in June 2026 to limit data exfiltration from prompt injections. The flaw stemmed from how popup windows were handled during HTML previews: nested sandboxed frames still permitted popups that escaped the sandbox, enabling silent chat data theft, phishing, and zero-click exfiltration in the desktop app; OpenAI fixed it on September 18, 2026.
Why it matters: Lockdown Mode was built to stop data leaving a conversation through malicious outbound requests, and a weakness in an unrelated application feature routed around that boundary entirely - a reminder that an AI control is only as strong as the application environment around it. The disclosure landed a week after Parameter Labs showed a separate bypass of ChatGPT's sandbox and CSP that let an attacker take over an account from a shared chat link.
Read the original at The Hacker News ↗
- Topics
- AI Security
- Related
- SECURITY AFFAIRS AI-CYBERSECURITY NEWSLETTER ROUND 8 Cyber Risks · Nov 15, 2026Pentagon Renames CDAO ‘Super Intelligence Office’ Cyber Risks · Nov 15, 2026186 Fake AI Domains Redirect Users to Click Fix Lures, Browser Malware, and Search Hijackers Cyber Risks · Nov 12, 2026