A Shared ChatGPT Link Could Run Attacker Code in Your Session and Read Private Chats

Security researchers at Parameter discovered a critical vulnerability in ChatGPT's shared link feature that allowed attackers to execute code within a victim's active session. Any logged-in user who opened a malicious ChatGPT share link was at risk of having their private conversations read by the attacker. The flaw was reported to OpenAI via their bug bounty program on October 18 and was patched eight days later. OpenAI classified the vulnerability as P1, their highest severity rating. The attack required no special privileges from the victim beyond simply clicking a shared link while logged in.