Shared ChatGPT link could run attacker code inside a logged-in session

A malicious ChatGPT share link could run an attacker's code inside the logged-in session of anyone who opened it, letting them read that person's private conversations, according to researchers at the security firm Parameter.

The researchers reported the flaw through OpenAI's bug bounty program on October 18, The Hacker News reported on October 31. OpenAI fixed it eight days later and rated it P1, the company's most serious category.

Why it matters

Share links are meant to be a low-risk way to pass a conversation to a colleague or friend. This case shows the opposite: opening a link someone else sent could have handed over code execution in the context of the victim's own session, with access to chats that were not meant to be shared. The report does not describe the technical mechanism, so it is not clear which parts of the session were reachable or what conditions an attacker needed.

What defenders should take from it

• Treat links to AI chat shares like any other untrusted link, especially when they arrive unsolicited.
• Keep AI assistant accounts on the same patching and monitoring footing as other SaaS tools, and watch for unusual session activity.
• Report suspected issues through the vendor's bug bounty channel rather than testing on other people's accounts.

The researchers have not published details of how the flaw worked. The report gives no indication that it was exploited in the wild.