Crafted ChatGPT share link could run code in a victim's session and read private chats until OpenAI's fix
Parameter's write-up says chaining the HTML Preview feature with a graph let attacker JavaScript run as chatgpt.com when a victim opened a public share, calling authenticated APIs to list and exfiltrate private conversations; OpenAI fixed it on 26 October and rated it P1. No exploitation in the wild is reported.