ChatGPT's New Locked Chats: Privacy Win or False Sense of Security?

OpenAI has introduced a new feature called Locked Chats for ChatGPT, designed to keep sensitive conversations private. You can lock a chat with a four-digit PIN or your face/fingerprint on mobile. Once locked, the chat disappears from your history and search results, so someone casually browsing your account can't see it. This is great for planning surprises or discussing private matters. But there are important caveats.

First, locking a chat doesn't instantly hide it everywhere. If you have the chat open in a browser tab, it may still be visible until you close or refresh the tab. Also, other devices where you're logged in might show the chat title until they refresh. And when you unlock a locked chat, it stays unlocked until you manually lock it again—so if you walk away, someone else could access it. You need to be diligent.

Second, locking a chat doesn't erase information you've already shared. If you created a shareable link to that chat, the link still works even after locking. You must delete the link separately. Also, if ChatGPT's memory feature picked up details from the conversation before you locked it, those details remain in its memory and can influence future responses. You have to manually remove them from the memory settings. The lock only prevents new memories from leaking into other chats after it's set.

For ultra-sensitive conversations, OpenAI suggests using Temporary Chat, which doesn't save to history, doesn't create memories, and isn't used for training. The trade-off is you can't return to the conversation later. Locked Chats are a useful step for everyday privacy, but they're not a comprehensive solution. You still need to manage shared links, memory, and open tabs to truly keep secrets under wraps.