ChatGPT Locked Chats: How PIN-Protected Hidden Chats Work

OpenAI has begun rolling out Locked Chats, a ChatGPT feature that lets you hide selected conversations behind a four-digit PIN or your device's own biometric unlock, starting with a gradual release on October 20, 2026, per OpenAI's ChatGPT release notes. Locked chats disappear from regular history, pinned chats and search, but OpenAI is explicit that locking is not encryption: the underlying data retention, model-training settings and legal obligations attached to a chat do not change just because it is hidden from view.

What ChatGPT Locked Chats Actually Is

Locked Chats is a visibility feature, not a new storage system. According to OpenAI's Help Center article on using locked chats in ChatGPT, a locked conversation is removed from your regular chat history, your pinned chats and your normal search results, and moved into its own section. You can find it by going to Projects, then Locked chats, and opening it with your PIN or supported device authentication. Titles for locked conversations are hidden in the sidebar, so even a glance at an unlocked phone or laptop will not reveal what a locked chat is about.

The point of the feature is straightforward: ChatGPT conversations increasingly include sensitive details — health questions, financial planning, legal concerns, relationship issues — and until now, anyone who could open your ChatGPT account (a shared laptop, a phone left unattended, a work device) could scroll through all of it. Locked Chats gives users a way to tuck specific conversations out of casual view without deleting them.

How to Lock a Chat in ChatGPT

Locking an eligible conversation takes a few steps, per OpenAI's documentation:

• Open the chat you want to hide and tap or click the three-dot menu (•••).
• Select "Lock chat" and confirm the action.
• The first time you lock a chat, you'll be asked to create a 4-digit account PIN. This PIN is tied to your OpenAI account as a whole, not to a single chat or workspace, so you only set it once.
• From then on, opening any locked chat requires that PIN, or an approved biometric method on supported devices.

To end an active unlocked session yourself, you can select "Lock now" from the same menu. OpenAI also builds in automatic re-locking: on the web, access ends as soon as you refresh the page or close the tab; on mobile, a locked chat you've opened will re-lock itself after about one minute away from the app or with the app in the background.

PIN, Face ID, Touch ID and Device Unlock

On web and desktop, the 4-digit PIN is currently the only way in — there's no biometric option on those platforms yet. On supported iOS and Android devices, you can additionally enable Face ID, Touch ID, or your phone or tablet's own screen-unlock method as a faster alternative to typing the PIN. Each device you use is configured for biometrics separately, and the PIN always remains available as a fallback if biometrics fail or aren't set up.

Changing your PIN later requires entering your current PIN first, even if you normally unlock with Face ID or Touch ID, and doing so does not affect any chats you've already locked — they stay locked and reopen with the new PIN. Desktop access to locked chats requires OpenAI's newer native ChatGPT app for macOS or Windows rather than a browser tab pointed at chatgpt.com. Voice mode works inside locked chats on web, iOS and Android, but not yet in the desktop app.

Which ChatGPT Plans Support Locked Chats

OpenAI is making Locked Chats available broadly rather than reserving it for paid tiers. The rollout, as described in OpenAI's own Help Center documentation, spans nearly every consumer and business plan:

Because the feature is rolling out gradually rather than flipping on for every account simultaneously, OpenAI's documentation notes that some users on an eligible plan may not see the "Lock chat" option in their three-dot menu immediately. If it's missing, it's likely a staged-rollout timing issue rather than a plan restriction, and it should appear as the rollout continues.

What Locking a Chat Actually Hides

Locking changes what's visible, in a few specific ways documented by OpenAI:

• The chat's title and full content disappear from your regular chat history, pinned chats list, and normal search results.
• In the sidebar, locked conversations may appear under a "Locked chats" project, but with their titles hidden rather than shown.
• Any files you uploaded to, or that were generated inside, a locked chat are hidden from the Space file view — including files created after you locked the conversation.
• Files that were added to a Space separately from the locked chat remain visible as normal; locking a chat doesn't retroactively hide unrelated files.

What Locked Chats Does Not Protect

This is the part OpenAI is careful to spell out, and it's the single most important thing to understand before relying on the feature for anything genuinely sensitive. Per OpenAI's Help Center article, locking a chat:

• Does not create end-to-end encryption or any kind of separate encrypted vault. The data isn't protected by a cryptographic key you control — it's a visibility layer inside the normal ChatGPT account system.
• Does not change OpenAI's data retention for that conversation. However long OpenAI would otherwise keep the chat, locking it doesn't shorten or extend that.
• Does not change your "Improve the model for everyone" setting. If that setting was on before you locked the chat, it stays on for that chat's data unless you change the setting itself.
• Does not remove information ChatGPT already remembered about you before you locked the conversation — memory formed earlier isn't retroactively erased.
• Does not override workspace policies, safety requirements, or legal obligations that would otherwise apply to the content.
• Does not stop anyone who already has access — to your unlocked device, your PIN, or your account — from copying the chat's contents once it's open.

In other words, Locked Chats is designed to stop casual viewing — a partner glancing at your screen, a colleague borrowing your laptop — rather than to provide cryptographic confidentiality against OpenAI itself, account compromise, or legal discovery.

Enterprise, Edu and Admin Controls

Organizations running ChatGPT Enterprise or Edu don't get an off switch for this feature. OpenAI's documentation states that workspace administrators cannot disable Locked Chats through any workspace setting or role-based permission, and admins cannot unlock a user's locked chats or reset a user's PIN on their behalf. At the same time, locking doesn't create a compliance blind spot: conversations locked by employees still appear in the conversation message logs available through OpenAI's Compliance Platform, so the content remains auditable at the organizational level even though it's hidden from the user's own everyday view.

Chats You Can't Lock

Locking isn't available for every kind of conversation. According to OpenAI, the "Lock chat" option won't work on:

• Temporary Chats — these have to be saved as a regular chat first before they're eligible
• Archived chats
• Group chats
• Conversations from the earlier ChatGPT Health experience (though chats using the newer Health app can still be locked)
• Chats created from a shared link
• Chats that have scheduled tasks attached, including paused ones (these need to be deleted first)
• Chats containing meeting recordings
• Chats tied to background tasks
• Chats whose responses were saved as a project source
• Codex conversations, including cloud Codex sessions — for more on how that product works, see our explainer on how GitHub Copilot and similar AI coding tools handle code review for a sense of how coding-focused chat sessions differ from everyday ChatGPT use
• Local Work conversations

Locking a chat that belongs to a project also removes it from that project entirely, along with the project's custom instructions, files and memory context — so a locked chat effectively becomes a standalone conversation. If a chat already has an active shared link, you can still lock it; the existing link keeps working, but no new messages sent after locking get added to what the link's viewers can see.

Locked Chats vs. Temporary Chat Mode

ChatGPT already had a privacy-oriented mode before this update: Temporary Chat. The two features solve different problems and aren't interchangeable, which is a common point of confusion.

Put simply: Temporary Chat is for a conversation you never want saved at all, while Locked Chats is for a conversation you want to keep — just not visible by default. OpenAI's own Temporary Chat FAQ covers that mode's rules on training data and the 30-day safety retention window in more detail.

How Memory Works Inside Locked Chats

Memory behavior around Locked Chats has its own set of rules, and they're a little more layered than the rest of the feature. Locked chats can draw on your regular Memory — so a locked conversation can still reference things ChatGPT already knows about you from earlier, unlocked chats. New information that ChatGPT would normally remember from messages sent after a chat is locked, however, is only reused in your other locked chats; it does not flow back into your regular, unlocked chats.

If you later permanently unlock a chat, the full conversation can go on to be used for regular Memory, provided your settings and applicable policies allow it — but any memories that were created specifically while the chat was locked are not automatically folded into your regular Memory just because you unlocked the chat. And if a workspace has turned Memory off entirely, locked chats in that workspace can't use or create memories either. For the broader mechanics of how ChatGPT's memory system works outside of locked chats, OpenAI's Memory FAQ is the authoritative reference.

What Happens If You Forget Your PIN

This is the feature's biggest single risk, and OpenAI states it plainly: forgetting your PIN, by itself, doesn't delete anything. But if you go through the process of resetting a forgotten PIN, that action permanently deletes every locked chat on your account, any memories that existed only inside locked chats, and any files or other content that was protected solely by the old PIN. OpenAI's documentation is unambiguous that this cannot be undone and that OpenAI Support has no way to recover what's deleted once a PIN reset goes through. Anyone planning to use Locked Chats for something they'd be upset to lose permanently should treat the PIN with the same care as a password they can't request a reset link for.

What's Next

Locked Chats arrives as ChatGPT keeps expanding what a single account can hold — projects, long-running Memory, scheduled tasks, Codex sessions — and as OpenAI continues rolling out interface changes alongside its broader model updates, including the GPT-6 rollout and its new interactive interface. A visibility-only privacy control for individual chats is a comparatively small addition next to those model changes, but it's one of the more directly useful ones for everyday account hygiene, particularly on shared or work devices.

What's still unclear is how far OpenAI will take this. The current feature set (PIN plus optional biometrics, no encryption, full compliance visibility for Enterprise and Edu) suggests OpenAI is treating this as a UI convenience rather than a security boundary. Whether a more robust, genuinely encrypted option ever follows — something closer to what some messaging apps offer — has not been indicated in OpenAI's own materials, and should be treated as an open question rather than an assumption. Since the rollout is still staged as of this writing, the most practical next step for most users is simply checking whether "Lock chat" has appeared yet in their own account's three-dot menu, and setting a PIN they won't forget before relying on it for anything they'd mind losing.

Written byTheo Park

Theo Park runs the AI desk at Pandromeda. He follows model launches from the frontier labs and the open-weight community, tracks the assistants and developer tools built on them, and explains what each release changes on pricing, capability and safety. His reporting leans on primary sources: model cards, technical reports, API documentation and the companies' own announcements.

More from Theo ParkAll AI coverage