ChatGPT locked chats hide conversations behind a 4-digit PIN

OpenAI has added locked chats to ChatGPT, according to its release notes entry dated October 20, 2026. Selected conversations disappear from history, pinned chats and search, and open only with a 4-digit account PIN or supported device authentication.

How locked chats work and who gets them

The entry appeared on OpenAI’s ChatGPT release notes page in the help center. It describes locked chats as a way to keep chosen conversations in your account while hiding them from the places where they normally show up: the regular chat history, pinned chats and search.

To lock a conversation, open the more options menu (the three dots) on a chat and select Lock chat. To read it again, you enter the PIN tied to your account, or use whatever device authentication your phone or computer supports. The notes do not say which methods count, such as fingerprint or face unlock.

OpenAI states that retention and model-improvement settings stay unchanged. In practice, a locked chat is treated like any other chat for storage and training purposes, and the lock only changes who can see it on a device.

The rollout is gradual and covers:

• Free, Go, Plus and Pro users
• ChatGPT on the web
• The iOS and Android apps
• The new ChatGPT desktop app for macOS and Windows

The notes list no Business, Enterprise or Edu plans, no regions and no date by which everyone will have it. Because the rollout is gradual, the option may not show up immediately on every account. OpenAI points to a separate help article on locked chats for more detail, which the release note links but does not reproduce.

A privacy step for shared screens, not for data

The feature addresses a familiar problem: ChatGPT conversations often contain health questions, money worries or drafts that users would not want a partner, colleague or anyone borrowing a laptop to see. Until now the options were deleting a chat, archiving it, or using a Temporary Chat, which is not saved to history. Archiving removes a conversation from the sidebar but, as far as public reporting shows, does not add a PIN or biometric check. Locked chats sit in between: the conversation persists, but it is gated.

The limits matter as much as the lock. OpenAI says retention and model-improvement settings are unchanged, so a locked chat is hidden from view rather than withheld from OpenAI’s systems. Anyone who has opted out of training keeps that opt-out, and anyone who has not is not newly protected. The release note does not say whether a locked chat can be shared, exported or surfaced through memory, and it does not describe what happens if a user forgets the PIN.

What to watch as the rollout widens

The feature did not arrive without warning. Before the release note, Android Authority reported finding references to a Locked Chats feature in version 1.2026.230 of the ChatGPT Android app. Those reports, which TechRadar also covered, described a lock by PIN or fingerprint and said new memories created in locked chats would be used only in other locked chats. The release note says nothing about memory, so that detail remains an unconfirmed claim from app code analysis, not an OpenAI statement. How memory behaves is the question most worth checking once the help article and the feature are live, because a hidden chat that quietly shapes answers elsewhere would defeat part of its purpose.

The timing also fits a pattern in recent OpenAI notes. In late September the company introduced security history, which lists sign-ins and changes to multi-factor authentication and passkeys, and it rolled out a Privacy Center to Free, Go, Plus and Pro users. Locked chats extend the same account-level controls to individual conversations. Meanwhile, OpenAI is pushing ChatGPT further into shared work: our earlier report on team tasks, Space and dots described a workspace where Pages and files are shared. That makes a personal, hidden layer more relevant, though locked chats are listed for consumer plans and not for those workspaces.

Three things would confirm how well it works. The first is whether the lock holds on shared devices, for example when someone else opens the desktop app on an unlocked computer. The second is how fast it reaches all four client types, since the rollout is gradual. The third is whether OpenAI extends it to Business and Enterprise accounts, where administrators set retention rules that may conflict with chats users can hide. The company’s claims about the feature have not been independently tested, and the release note is the only official description so far.