OpenAI adds locks to ChatGPT conversations without changing data controls

ChatGPT users are getting a way to keep selected conversations out of sight without giving them up. OpenAI began rolling out Locked Chats on October 20, 2026, protecting conversations with a four-digit PIN or supported device authentication. The feature limits access when someone else uses a device; it does not change how long those conversations are retained.

Keep the conversation, hide it from ordinary browsing

OpenAI describes a familiar situation behind the feature: someone hands a phone to a friend or family member to show one conversation, while another contains something personal. Planning a surprise is another intended use. The lock is designed to keep those selected exchanges private during that kind of device sharing.

To lock a conversation, open its more-options menu, shown as three dots, and select Lock chat. The conversation moves into a dedicated Locked Chats section. It no longer appears in regular chat history, pinned conversations or chat search results, separating it from the places someone would normally browse.

Opening that section requires the account’s PIN or supported device authentication, including Face ID or Touch ID. Locked Chats remain available to return to later. That is the practical distinction from the separate sessions provided by Temporary Chats: this feature is built around keeping a conversation, rather than making it temporary.

New memories stay inside the locked section

The separation extends beyond the conversation list. When Memory is enabled, Locked Chats have their own memory, shared across conversations inside that section. New messages sent after locking can update this separate memory, but they do not update regular Memory. Information added there therefore stays out of the memory used in ordinary chats.

That boundary applies to new messages, not everything previously discussed. Locking a conversation does not remove information ChatGPT remembered before the lock was applied. Moving an existing chat into the protected section is therefore not a way to undo its earlier contributions to regular Memory.

OpenAI also says the feature does not create end-to-end encryption or a separate encrypted vault. The “Improve the model for everyone” setting stays as it was. Choosing which conversations require authentication and choosing whether conversations may be used for model improvement remain separate decisions; applying a lock does not change the latter.

Access is gradual, and some chats are excluded

OpenAI’s October 20 release note says Locked Chats are rolling out gradually to Free, Go, Plus and Pro users. It names the web, iOS and Android apps, and the ChatGPT desktop app for macOS and Windows. Eligible users may not see the option immediately as the rollout proceeds.

TechCrunch also lists Business, Enterprise and Edu among the plans that will have access, citing OpenAI’s guidance. ChatGPT for Healthcare is excluded. That broader plan list goes beyond the individual subscriptions named in the dated release note; it should not be read as a guarantee that every supported account already has the feature.

Even with access, users cannot lock every conversation type. The exclusions cover:

• Temporary and archived chats.
• Group chats and chats created from a shared link.
• Chats containing scheduled tasks or meeting recordings.