# ChatGPT Locked Chats: How to Use a PIN and What Stays Private

OpenAI now lets you hide saved conversations behind a PIN, but locking a chat leaves several important data controls unchanged.

By Paras Madan | 2026-10-21


OpenAI began rolling out locked chats on October 20, 2026. The feature keeps selected ChatGPT conversations out of ordinary history and search, then requires a PIN or supported device authentication to open them. It is useful for screen sharing and shared devices, but it does not create an encrypted vault.

The distinction is worth understanding before moving sensitive conversations. Hiding a chat from someone looking at your screen and changing how a service handles its data are separate controls.

## Who can use ChatGPT locked chats?

The [ChatGPT release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes) confirm a gradual rollout to Free, Go, Plus and Pro on the web, iOS, Android and the new desktop app for macOS and Windows. [Business](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes), Enterprise and Edu are also supported. The feature excludes ChatGPT for Healthcare, local Work and Codex conversations.

If Lock chat is missing, check rollout availability and conversation eligibility before assuming the feature is broken.

## How to lock and reopen a chat

Open an eligible conversation's three-dot menu, select **Lock chat** and confirm. On first use, create a four-digit account PIN. To return later, open Projects, then Locked chats, and authenticate.

OpenAI's [locked-chat guide](https://help.openai.com/en/articles/20001411-using-locked-chats-in-chatgpt) explains temporary access and relocking. Select Lock now when finished.

Refresh other open devices before screen sharing. Previously loaded content can remain visible, and locking a web conversation can leave it open in that tab.

## What locking changes, and what it does not

**Locked chats are an access and visibility control.** They do not add end-to-end encryption, change retention or switch off the model-improvement setting. Information remembered before locking is not automatically removed.

A previously created shared link is another separate boundary. Locking the conversation does not revoke access to content already shared through that link.

For organizations, the [Enterprise and Edu release notes](https://help.openai.com/en/articles/10128477-chatgpt-enterprise-and-edu-release-notes) explicitly say retention policies still apply and locked conversations remain in message logs available through the OpenAI Compliance Platform. A hidden sidebar entry should never be mistaken for an exemption from workplace data rules.

## Files follow their original access rules

OpenAI's [file-management documentation](https://help.openai.com/en/articles/20001052-using-library-to-manage-files-in-chatgpt) distinguishes files created or uploaded in the locked conversation from files added from elsewhere.

Protected files can be opened only in their originating locked chat, not reused in another conversation while locked. Files brought in from elsewhere or from connected services retain their existing access rules. Permanently unlocking restores access to protected files that had previously been saved outside the conversation.

The help pages currently use both Library and Space for these file surfaces. The consistent rule is the file's origin and access boundary, rather than the navigation label.

That matters when a conversation includes a document already shared or stored elsewhere. Locking one copy of the discussion does not change every other place that document exists.

## Be careful with a forgotten PIN

Changing a PIN you know preserves the chats. Resetting a forgotten PIN permanently deletes all locked chats and content protected only by the old PIN, including the relevant files and memories. OpenAI says Support cannot recover that content.

This makes PIN management part of the feature, rather than a minor setup detail. Before using locked chats for anything important, decide whether you need protection from casual viewing, different data-use settings, or both. Each requires its own check.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help AI teams explain product changes clearly, including the details that determine how people actually use them.

[Book a 30-minute call](https://cal.com/paras-madan/30min).