This article was translated from Arabic. Read the Arabic original
Austria Now News – Austria
OpenAI, the developer of ChatGPT, has closed a serious security flaw that allowed strangers to view other users’ chats via manipulated share links, oe24 reported, citing the IT security portal The Hacker News.
Highest severity level
According to the report, the company classified the problem at its highest priority level, P1. Those affected were users who unknowingly opened a pre-prepared share link to a chat history.
The flaw was discovered by experts at the IT security company Parameter, who reported it to OpenAI.
How did the flaw work?
According to the report, the manipulated link allowed foreign code to run within the user’s active session, enabling the researchers to read unpublished chats, among other things. The researchers said they could have carried out practically any action in the name of the victim’s ChatGPT account.
The report stated that the test was conducted exclusively in a controlled test environment. The flaw was caused by faulty processing of browser messages in a mathematics library used by the program, which made it possible to bypass the web application’s protection and inject the malicious code via a trusted interface.
No action needed from users
According to The Hacker News, citing Parameter, OpenAI responded quickly and fixed the problem on its servers within eight days. ChatGPT users therefore do not need to take any security measures themselves.
Austria Now News – always keeping you up to date.
Symbolic image
Sources
- oe24, Manipulierte Links legen private ChatGPT-Chats offen, 01.11.2026, oe24.at
