OpenAI has rolled out a privacy-minded update for ChatGPT that lets users hide specific conversations behind a PIN or supported biometric lock. The new Locked Chats feature is built for people who want to keep sensitive conversations out of plain view without deleting them, while still being able to return to them later.
That matters in a world where AI chats often wander into personal territory fast. People use ChatGPT for health questions, family issues, gift planning, money worries, and plenty of other moments they would rather not have sitting in full view on a shared device.
The basic idea is simple: select a chat, lock it, and keep it behind a four-digit PIN. On compatible phones, users can also unlock conversations with Face ID, Touch ID, or the device’s built-in authentication method.
Locked Chats does more than just hide a thread from casual browsing. Once a conversation is locked, it moves out of the standard history and into a separate area, so it is no longer visible in the usual list of chats or basic search results.
There is a catch, though, and it is an important one. If a conversation was part of a Project, locking it removes it from that space, which means it also loses access to that project’s instructions, files, and memory.
The feature is rolling out across many ChatGPT plans, including Free, Go, Plus, Pro, Business, Enterprise, and Edu. It works on the web, on iPhone and Android, and in the newer desktop apps for Mac and Windows, although some versions and account types have limits.
Getting a conversation under lock does not require much effort. A user opens the chat, taps the three-dot menu, selects Lock chat, confirms it, and then creates a four-digit PIN if one has not already been set up.
After that, the conversation shifts into Locked chats, where it stays tucked away until opened again. The same PIN can be used on supported devices tied to the same OpenAI account, which keeps the process consistent across platforms.
Biometric access is where the feature starts feeling especially convenient. Instead of typing a PIN every time, compatible iPhones can use Face ID or Touch ID, while supported Android phones can use device unlock methods such as fingerprint authentication.
That convenience does not mean the protection is identical everywhere. The authentication method depends on the phone, the app version, and the device settings already in place, so the exact experience may vary from one person to the next.
Opening a locked chat later is straightforward too. Users go back into Projects, open Locked chats, enter the PIN or authenticate with the device, and then choose the conversation they want to view or continue.
One useful detail is that new messages can still be added inside a locked conversation. The thread stays protected while you continue talking, which makes it handy for an ongoing topic that still feels too personal to leave out in the open.
Still, the privacy picture is not as sealed as the word “locked” might suggest. OpenAI says Locked Chats are not end-to-end encrypted, do not create a separate encrypted vault, and do not change how long the company retains information.
Memory also deserves a close look. Older remembered details may still live in regular Memory, while new information from locked conversations stays tied to the locked area, which means the feature controls visibility more than it erases all trace of context.
Shared links are another weak point. If a conversation was already shared through a link, locking it does not cancel that access, so anyone with the link may still be able to see what was previously shared unless the link itself is removed.
Files follow a similar pattern of partial protection. Anything uploaded or created inside the locked conversation gets the same protection as the chat, but files stored elsewhere in ChatGPT or connected services keep their own permissions and do not automatically become private just because the chat is locked.
There is also an important warning for workplace accounts. Enterprise and Edu systems may still retain logs through compliance tools, and administrators cannot reset a PIN or unlock chats, but organizational policies can still apply to the data.
For people who want a different kind of privacy control, ChatGPT also offers Temporary Chat. That option keeps conversations out of the normal history and avoids creating new memories, but it is designed for one-off discussions rather than a saved thread you want to revisit later.
Locked Chats make the most sense when a conversation matters enough to keep, but not enough to leave exposed. It gives users a little breathing room when AI is handling topics that are personal, awkward, or just plain private, without forcing them to delete the whole exchange.
Even then, the feature has a sharp edge: forgetting the PIN can be costly. Resetting it permanently deletes locked chats and any content tied only to that protected space, and OpenAI says that loss cannot be undone.
That makes the feature feel useful, but not magical. It adds a real layer of everyday privacy, yet the smartest move is still to think twice before putting deeply sensitive information into any chat, locked or not.
