Breaking

OpenAI Patches ChatGPT Flaw That Let Malicious Shared Links Run Attacker Code and Expose Private Chats

October 31, 2026 at 01:35 EDT

Security

OpenAI has fixed a vulnerability in ChatGPT that could let a malicious shared link execute attacker-supplied code inside a user's logged-in session and steal private chat content, according to security researchers who demonstrated the attack.

Security · ChatGPT

A shared link could run attacker code in a logged-in ChatGPT session

OpenAI has fixed a flaw in how ChatGPT handled shared links. Researchers demonstrated that a crafted link could execute attacker-supplied code inside a user's authenticated session and steal private chat content.

How the attack chain worked

Trust in links and session handling, not the AI model's outputs

STEP 1

Crafted link

A user opens a malicious shared link.

→

STEP 2

Code runs

Attacker code executes inside the user's authenticated session.

→

STEP 3

Chats read

The session grants access to the user's own conversations.

→

STEP 4

Content exfiltrated

Private chat content is sent out, without obvious user signs.

Why it matters

A single click can compromise an account, so security teams treat this class of flaw as high priority.

Business data: chats may contain internal information.
Personal details: private information can be exposed.
Source code: technical material can be read by attackers.

What users can do

Avoid opening unfamiliar shared links.
Review active sessions in account settings.
Sign out of shared devices.

Still not public

How many users may have been affected.
Whether exploitation occurred before the fix.
The exact timeline from discovery to patch.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.