OpenAI has paid a $5,000 bug bounty for a ChatGPT share link flaw that could let an attacker run code inside the logged-in sessions of everyone who opened the malicious link, according to a report by Swati Khandelwal in The Hacker News and surfaced on Techmeme. The bug is a stark reminder that shared AI chats are not just passive text.
The problem matters because share links are a normal part of ChatGPT use, and users often open them while signed in. In this case, any script hidden in a shared session could reach the viewer's own account. OpenAI has addressed the issue with the bounty, though details on when the fix landed remain limited.
How the ChatGPT share link flaw worked
According to the report, the shared session acts as a delivery vehicle: attackers create a ChatGPT share link, inject code, and wait for logged-in users to open it. Because the browser trusts the active session, the code can act with the viewer's permissions.
That is a classic cross-session risk: the user never downloads anything or shares a password, yet their session is exposed simply by clicking a link that looks like normal AI output. Researchers earned $5,000 for reporting the flaw through OpenAI's bug bounty program.
OpenAI subscription limits arrive alongside the fix
In a separate policy change, The Decoder reports that OpenAI is now capping the number of paid ChatGPT accounts tied to a single device and payment card. Paid plans including ChatGPT Go, Plus, Pro 100, Pro 200 and Pro 500 are affected, while Free, Business, Enterprise and Edu accounts are excluded.
Users who reach the cap see a 'Device limit reached' message, and that account can no longer be used on the device. Logging out does not free up a slot, and OpenAI points users toward extra credits or usage resets rather than stacking cheap subscriptions.
