Affected Systems
OpenAI ChatGPT with Lockdown Mode enabled, particularly the Electron-based desktop application. Vulnerability patched September 18, 2026. Affects users who enabled the optional Lockdown Mode security setting introduced in June 2026.
Exploitation Status
Patched vulnerability. No evidence of active exploitation in the wild. Proof-of-concept demonstrated by Check Point researchers. Exploitation requires attacker to deliver crafted prompt, malicious GPT, or shared chat link to victim.
Business Impact
Organizations using ChatGPT for sensitive workflows faced risk of silent data exfiltration from chat sessions, bypassing the Lockdown Mode protection designed to prevent prompt injection attacks. Attack vectors included malicious shared chats, custom GPTs, or crafted prompts. Desktop application users faced zero-click exfiltration risk. Vulnerability also enabled phishing and malware delivery through convincing fake update prompts. No CVE assigned. Patch deployed September 18, 2026.
Urgency
🟡 Within a week
Recommended Actions
- Verify all ChatGPT desktop and web clients are updated to versions released after September 18, 2026
- Review organizational policies on accepting shared ChatGPT links and custom GPTs from untrusted sources
- Audit ChatGPT usage logs for suspicious shared chat interactions or custom GPT installations between June and September 2026
- Reinforce user training on verifying authenticity of update prompts and download requests within AI applications
- Evaluate whether Lockdown Mode remains appropriate security control for sensitive use cases given historical bypass vulnerabilities