ChatGPT share links could hijack accounts until an October fix

Security firm Parameter says it found a bug where opening an attacker's public ChatGPT share link could run the attacker's code inside the victim's signed-in ChatGPT session. In its proof, the code listed the victim's conversations, read a private chat that was never shared, and sent part of it to an outside server. The chain combined ChatGPT's HTML Preview, a maths-keyboard component loaded by an interactive graph, and a YouTube script address allowed by ChatGPT's security policy. Parameter reported it on 18 October; OpenAI rated it P1, confirmed the fix on 26 October and paid a $5,000 bounty. On Hacker News (23 top-level comments when we read it) the loudest theme was that $5,000 is too low for a P1 bug, a couple of commenters criticised Parameter for testing their own systems unasked, and one praised the step-by-step write-up; no consensus beyond the bounty complaints. Sorted take: nothing to patch on your side, but your chat history is only as safe as the account.