ChatGPT’s new Locked Chats feature removes selected saved conversations from ordinary history and search, then puts them behind a PIN. Users can keep sensitive discussions out of view when someone else has access to their screen or device.
OpenAI began gradually rolling out the feature on October 20, 2026, according to TechCrunch’s report. Users can protect locked conversations with a four-digit PIN, with Face ID or Touch ID also available on supported devices.
The protection addresses someone browsing an already accessible ChatGPT account. It does not create end-to-end encryption, establish a separate encrypted vault, or change OpenAI’s retention and model-training settings.
Recovery comes with a consequential warning: resetting the PIN permanently deletes all locked chats, files and other content protected by the previous PIN.
Locked Conversations Leave History and Search
Once a conversation is locked, it moves into a dedicated Locked Chats section. OpenAI says these conversations no longer appear in regular chat history, pinned conversations or chat search results.
Removing those entry points makes a sensitive conversation less likely to appear while someone browses the sidebar or searches for an unrelated discussion. The feature is designed for situations such as handing a device to a friend or family member to show them something in another chat.
Locked Chats remain saved conversations that users can return to. Temporary Chats, by comparison, provide a separate session that does not appear in chat history. Locking restricts access to a conversation you intend to keep.
Access requires a four-digit PIN or, on supported devices, Face ID or Touch ID. Biometric support does not mean every device will offer the same authentication method.
This is an additional access barrier inside ChatGPT. It does not replace the need to secure the device or account itself, and removing a conversation from ordinary search does not delete it.
Separate Memory Is Shared Within the Locked Section
With Memory enabled, Locked Chats has its own separate memory shared across conversations inside that section. New messages sent after a conversation is locked can update this locked-chat memory without updating regular ChatGPT memory. Information newly supplied there is not supposed to become part of the regular memory used outside that section.
Before moving an existing conversation, there are two qualifications to understand.
First, memory is shared across the locked section. Each locked conversation does not receive its own isolated memory compartment. Someone who wants to keep two sensitive subjects separate should not assume that placing both conversations behind the PIN also separates their remembered context.
Second, locking a conversation does not erase information ChatGPT remembered before it was locked. OpenAI explicitly identifies this as a limit of the feature.
Moving an existing personal discussion into Locked Chats, for example, does not retroactively clean up regular memory. If ChatGPT already remembered information from that discussion, locking the source conversation leaves that memory in place. The feature controls new memory updates without automatically undoing previous ones.
Resetting the PIN Deletes the Protected Content
OpenAI’s most important warning concerns resetting the PIN: doing so permanently deletes all locked chats, files and other content protected by the previous PIN.
The consequence applies to the entire collection protected by that PIN, beyond any single conversation. It differs from the familiar account-password reset, where choosing a new credential normally restores access to existing material. OpenAI describes a destructive reset for Locked Chats.
Before putting an important saved discussion behind the lock, consider whether you could afford to lose it, particularly if it contains information you have not preserved elsewhere.
Keeping another copy introduces its own privacy tradeoff: a copy outside Locked Chats would need separate protection. The sensible preparation depends on the material. In any case, do not initiate a reset expecting the previous locked content to reappear afterward.
This deletion behavior does not establish anything about the feature’s encryption architecture. OpenAI expressly says Locked Chats does not use end-to-end encryption.
Locking Does Not Change Retention or Training Settings
OpenAI says locking a chat does not create end-to-end encryption or a separate encrypted vault. It also does not change how long the chat is retained.
Locked Chats controls whether someone can open a conversation through the normal ChatGPT interface without passing the additional access check. It establishes no new policy for how the service stores or handles that conversation.
Locking a conversation also leaves the Improve the model for everyone setting unchanged. Users should not interpret the lock as an additional training opt-out.
A user concerned about someone casually opening a personal conversation on a shared device gets a concrete control. Someone seeking a new retention policy or an end-to-end encrypted place to store sensitive material will not find that here.
For an existing conversation, three separate checks may matter: whether it is locked, whether regular memory already contains information from it, and what data controls apply to the account. Changing the first does not automatically change the other two.
Some Chat Types Cannot Be Locked
TechCrunch reports that the Locked Chats option is unavailable for:
- Temporary chats.
- Archived chats.
- Group chats.
- Chats created from a shared link.
Sources
- TechCrunch’s reporttechcrunch.com
