Skip to content
NewsProduct launch

OpenAI adds PIN-locked chats to ChatGPT, but skips encryption

· by Pondero Newsdesk

The short version

OpenAI began rolling out Locked Chats in ChatGPT on October 20, 2026, hiding conversations behind a four-digit PIN or Face ID, while the company says the feature does not encrypt messages or change how long it keeps them.

OpenAI adds PIN-locked chats to ChatGPT, but skips encryption

ChatGPT's new PIN lock keeps a friend holding your phone out of a sensitive conversation, but OpenAI says the chat itself stays exactly as readable to OpenAI, and exactly as remembered by the model, as it was before you locked it, per TechCrunch's report on the rollout.

What

OpenAI said on Tuesday, October 20, 2026, that users can lock individual ChatGPT conversations with a four-digit PIN, or with Face ID or Touch ID on supported devices, per TechCrunch. OpenAI described the use case as discussing something personal, or planning a surprise, then handing a phone to a friend or family member to show them a different chat without that other conversation being visible. A locked chat moves into a dedicated "Locked Chats" section and disappears from regular chat history, pinned conversations, and search results until the PIN unlocks it again. That is a different design from ChatGPT's existing Temporary Chats, which never get saved to history at all; a locked chat is saved, and a user can come back to it anytime after entering the PIN.

OpenAI was explicit about what the feature does not do. Locking a chat does not add end-to-end encryption or create a separate encrypted vault. Retention stays the same, the "Improve the model for everyone" training setting is unaffected, and nothing ChatGPT already remembered before the lock gets erased. With Memory turned on, a locked chat keeps its own memory shared only across other chats inside that same locked section; new messages sent after locking update that separate memory rather than a user's regular Memory, so new information from a locked chat does not carry over into unlocked chats. Resetting the PIN permanently deletes every locked chat, file, and other item tied to the old PIN, with no recovery option described. The feature is not available on temporary chats, archived chats, group chats, chats opened from a shared link, chats with scheduled tasks, or chats with meeting recordings, and it is rolling out to Free, Go, Plus, Pro, Business, Enterprise, and Edu plans, but not to ChatGPT for Healthcare. Because rollout started Tuesday, OpenAI said the option may not be visible to every account yet.

The PIN hides chats from a screen, not from OpenAI's servers

For anyone using ChatGPT on a shared laptop or a phone they hand to a kid or a coworker, Locked Chats closes a real gap: a specific conversation can no longer be stumbled on by someone scrolling the sidebar. It does not close the gaps that matter for anyone worried about what OpenAI itself can see, how a locked chat might surface in a legal discovery request, or how long the underlying data sits on OpenAI's servers, since none of those change when a chat is locked. Teams handling anything genuinely sensitive, client details, health information outside Healthcare plans, unreleased business terms, should treat the lock as a screen-privacy control for their own device, not as a security boundary against OpenAI or against a subpoena.

The PIN-reset behavior is the other operational detail worth flagging before anyone relies on this. Because a reset wipes every locked chat with no stated way to recover them, locking something important and then forgetting the PIN is equivalent to deleting it. Anyone who locks a chat they intend to keep should treat the PIN the way they would treat a disk-encryption passphrase, not a login password they can reset without consequence.

The walled-off memory has a trade-off too. A locked chat only draws on and updates the separate memory shared across the Locked Chats section, so anything discussed there stops informing ChatGPT's regular answers in unlocked chats. A manager who locks a recurring HR or performance conversation keeps it out of casual view, but also stops ChatGPT from carrying that context into the unlocked chats where most day-to-day work happens, a trade worth weighing before locking anything meant to be an ongoing reference.

Context and reactions

TechCrunch's report framed the launch against ChatGPT's growing use for personal matters well beyond coding or writing help, including relationship and dating advice, citing the BBC's reporting on that trend. TechCrunch also noted directly that while locking a chat can limit who sees it on a device, "it doesn't address concerns about how sensitive information is stored or handled," a distinction OpenAI's own feature description backs up point for point. The company has not said whether end-to-end encryption for locked or any other chats is planned.

What to watch next

Whether OpenAI follows with true end-to-end encryption for locked chats, a gap the launch leaves open, and whether the exclusion list (group chats, shared-link chats, meeting-recording chats) narrows as the feature matures past its initial rollout.

Sources