OpenAI adds PIN and biometric locks for sensitive ChatGPT conversations

How Locked Chats function

OpenAI is introducing a new privacy layer called Locked Chats. This tool allows users to select specific conversations and hide them from the standard interface. When a chat is locked, it is removed from the regular chat history, the pinned chats section, and the standard conversation search.

To access these hidden threads, users must navigate to the Projects section and select Locked chats. The system requires a four-digit account PIN to view the content. For users on mobile devices, OpenAI is integrating native security options, allowing for access via Face ID, Touch ID, or the device's standard unlock method.

The feature is designed primarily for situations where a user might be sharing their screen or using a device that other people can see. By locking a chat, the user ensures that sensitive topics do not pop up unexpectedly during a demonstration or a casual browse of the chat history.

What stays the same

While the feature adds a layer of visual privacy, it is important to understand what it does not do. As reported by Social Samosa, Locked Chats do not introduce end-to-end encryption. This means that while the conversation is hidden from your screen's history, the underlying data is still subject to OpenAI's standard chat retention policies.

The tool also does not change how the AI learns from your interactions. It will not alter your existing model-improvement settings. Furthermore, any information that ChatGPT had already 'remembered' before the conversation was locked will remain in the system's memory. The lock acts as a gate for the user's view, rather than a total wall for the data itself.

Additionally, the feature has specific limitations regarding what types of chats can be secured. Users cannot lock temporary chats, archived conversations, group chats, or conversations that were created from shared links. These types of interactions remain visible in the standard history regardless of the new tool.

Files and workspace rules

The security of Locked Chats extends to the files uploaded or created within those specific threads. When a conversation is locked, these files are hidden from the 'Space' area and search results. However, if a file was added to a Space separately from a locked conversation, it will keep its original access permissions.

For users in professional environments, there are specific rules for ChatGPT Enterprise and Edu workspaces. In these settings, locked conversations are still included in message logs that are accessible through OpenAI’s Compliance Platform. This means the feature is intended for individual privacy rather than bypassing company-wide oversight or workspace policies.

Users should be cautious when managing their security keys. If a user forgets their four-digit PIN and chooses to reset it, the system will permanently delete all locked chats. This includes any information the AI remembered only within those chats and any files protected by the old PIN. OpenAI Support cannot recover any content lost during this reset process.

Availability and setup

OpenAI is making this feature available across almost all subscription tiers, including the Free, Go, Plus, Pro, Business, Enterprise, and Edu plans. On desktop computers, the feature is supported specifically through the new ChatGPT app for macOS and Windows.

To set up a locked chat, a user can open the three-dot menu on an eligible conversation and select 'Lock chat.' The first time this is done, the user will be prompted to create their PIN. This PIN is universal across all supported devices signed into the same OpenAI account, providing a seamless experience for users who switch between phones and computers.

Because the rollout is happening gradually, some users may not see the option immediately. Once a conversation is unlocked, it returns to the regular chat history, though it may take some time for it to reappear in the search results. This gradual deployment allows OpenAI to monitor how the feature interacts with different user workflows and device types.