ChatGPT’s Locked Chats Keep People Out of Your Screen. Not Out of Your Data.
TL;DR: OpenAI began rolling out locked chats on 20 October — put a four-digit PIN, Face ID or Touch ID on individual ChatGPT conversations and they vanish from your history and search. It is a good feature for the problem it solves: someone else holding your phone. It is not encryption, it does not change what OpenAI keeps, and on a work account your locked chats are still in the compliance logs.
What locked chats actually do
Lock a chat from its three-dot menu and it moves out of your history, your pinned chats and your search results, into a separate section. The first time you do it you create a four-digit PIN for the whole account, not for that chat. On an iPhone, Face ID or Touch ID can stand in for it; on Android, your device unlock does.
The details worth knowing, which Notebookcheck set out from OpenAI’s own help documentation:
- On mobile, access expires after one minute in the background. In a browser, reloading or closing the tab re-locks everything.
- Android blocks screenshots of locked chats.
- Locked chats keep their own Memory, which does not feed back into your normal conversations.
- You cannot lock temporary, archived or group chats, chats with scheduled tasks, or chats created from a shared link.
- Available on Free, Go, Plus, Pro, Business, Enterprise and Edu — not ChatGPT for Healthcare — and rolling out gradually, so your account may not have it yet.
One warning before you touch it. If you forget the PIN, the only route back is a reset, and a reset permanently deletes every locked chat, the files inside them, and any memories stored only in locked chats. OpenAI’s help documentation, as quoted by Notebookcheck, is blunt about it: “OpenAI Support cannot recover the deleted content.” A workspace administrator cannot reset it for you either. Write the PIN down somewhere that is not ChatGPT.
Five things locked chats don’t protect
This is the useful part, and most of the coverage skipped it.
- It is not encryption. OpenAI says — in wording both outlets report from its help pages — that locking does not “create end-to-end encryption or a separate encrypted vault.” The conversation is stored exactly as it was; the lock governs who can open it on your screen.
- It does not change retention. Locking a chat does not shorten how long OpenAI keeps it.
- Training is unaffected. If “Improve the model for everyone” is switched on for your account, locked chats can still be used for it. The lock is not a consent setting.
- On Enterprise and Edu, the logs still have it. Locked chats remain in the OpenAI Compliance Platform message logs. If you are locking chats on a work account to keep them from your employer, that is not what this does.
- Old memories stay put. Anything ChatGPT already stored in Memory before you locked the chat is still in Memory, usable in ordinary conversations.
There is a sixth, almost too obvious to list: anyone who can open a locked chat can copy what is inside it. And if your account shows ads, ads appear in locked chats too.
Why this matters if you don’t code
We have now written three of these in a week, and the pattern is the point.
On Tuesday, Meta suspended Muse’s chat sharing after researchers found shared conversations were discoverable by people who were never sent the link. The week before, we looked at what Claude Dashboards can actually see. Now ChatGPT ships a lock that stops a person and not a system.
Each of these is a reasonable feature. The trouble is the words. “Share”, “lock”, “private” are borrowed from filing cabinets and front doors, and they describe who can look at your screen. The thing most people actually want to know is who keeps a copy — and that is a different question with a different answer in every tool.
So the practical test for any privacy control in an AI tool is to ask it twice. Who does this stop? Then: what does the company still have? Locked chats answer the first question well. They do not change the answer to the second one at all, and OpenAI, to its credit, says so plainly in its own help pages rather than letting the name do the implying.
Who should care (and who shouldn’t)
- Sharing a device with family: this is exactly what it is for. Turn it on, write down the PIN.
- Running a small business: useful for the screen-over-shoulder problem, useless for anything you would not want on the company record.
- On a work Enterprise or Edu account: read point 4 twice. Locked chats are still in your organisation’s compliance logs.
- Worried about training data: this is the wrong control. Turn off “Improve the model for everyone” instead, or use a temporary chat.
- Learning or building a prototype: no action needed. Just do not assume a lock makes a conversation safe to put anything in.
Our take
We would turn it on, and we would be clear with ourselves about what it buys. The design decisions are good ones — a one-minute background timeout, screenshot blocking on Android, a separate Memory so your locked conversations do not leak into everyday replies. Those are the marks of a team that thought about the actual threat, which is a person holding your unlocked phone.
What we would not do is let locked chats change what we are willing to type. The lock does not make a conversation disappear from OpenAI, from a compliance log, or from a training set you have already consented to. If a conversation would be a problem for you in any of those places, a PIN is not the fix — not sending it is.
Not sure which AI tool you should be trusting with what? Take the 60-second Vibe Coding Tool Finder quiz →
Some links on this page are affiliate links. If you buy through them we may earn a commission at no extra cost to you.
FAQ
Are locked chats in ChatGPT encrypted?
No. OpenAI says locking does not create end-to-end encryption or a separate encrypted vault. The conversation is stored the same way it always was; the PIN controls who can open it in the app, not how it is held.
What happens if I forget my ChatGPT PIN?
Resetting it permanently deletes every locked chat, the files in them, and any memories stored only in locked chats. OpenAI’s documentation says support cannot recover the deleted content, and a workspace admin cannot reset it on your behalf.
Can my employer still see my locked chats?
On Enterprise and Edu workspaces, yes — locked chats remain in the OpenAI Compliance Platform message logs. Locking hides a conversation from someone looking at your screen, not from your organisation’s records.
