Skip to content
Vulnerability in ChatGPT Lockdown Mode Enables Silent Data Exfiltration

Vulnerability in ChatGPT Lockdown Mode Enables Silent Data Exfiltration

First seen 5 Nov 2026, 18:47 UTC • •
API

Article Content

Browse articles
ThreatCluster AI
ThreatCluster •November 6, 2026 at 11:19 UTC

Researchers from Check Point disclosed a vulnerability in OpenAI's ChatGPT Lockdown Mode, which was designed to prevent data exfiltration through prompt injections. This flaw, found in HTML previews, allowed attackers to bypass security measures and exfiltrate chat data silently, enabling phishing and malware delivery without user interaction. The vulnerability was patched by OpenAI on September 18, 2026, following responsible disclosure. The issue highlights broader challenges in AI security, as the vulnerability stemmed from the application environment rather than the AI model itself. The attack could exploit multiple vectors, including crafted prompts and shared chats, affecting users of the ChatGPT desktop application. OpenAI's Lockdown Mode was intended to limit potential data leaks, but this vulnerability revealed an alternative path for data exfiltration. Check Point confirmed that the technique used to exploit this vulnerability is no longer effective after the patch.

Start a free Starter trial for enhanced analysis

Ask AI about this cluster

Updated 3d ago How this analysis works

Timeline

2026-09-18
Vulnerability patched by OpenAI
OpenAI addressed the vulnerability in ChatGPT's Lockdown Mode following responsible disclosure from Check Point Research.
Thehackernews
2026-11-05
Vulnerability disclosed
Check Point Research revealed the vulnerability that bypassed Lockdown Mode, allowing silent data exfiltration.
Blog.Checkpoint

More articles in this cluster (12)

Following this threat?

Track OpenAI in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.

Free account · no card needed

Extracted Entities (10)

1 / 3

Common questions

What was the nature of the vulnerability?
The vulnerability allowed attackers to bypass Lockdown Mode protections in ChatGPT, leading to potential silent data exfiltration.
How was the vulnerability exploited?
Attackers could exploit the vulnerability through HTML previews, enabling phishing and malware delivery without user interaction.
Is the vulnerability still active?
No, OpenAI patched the vulnerability on September 18, 2026, and it is no longer exploitable.