ChatGPT Adds Locked Chats for Private Conversations

Overview

OpenAI has introduced a new Locked Chats feature to ChatGPT, allowing users to secure sensitive conversations behind additional authentication layers. The update addresses growing privacy concerns as the platform is increasingly used for personal and confidential discussions. Users can now protect specific threads using a four-digit PIN, or leverage biometric authentication like Face ID or Touch ID on compatible devices.

This addition marks a significant step in enhancing user privacy within an ecosystem that often handles highly personal data. The feature is designed for scenarios where users may need to hand their device to others while keeping certain topics strictly private. It separates these protected conversations from the general chat history, ensuring they remain inaccessible to casual observers.

How Locked Chats Work

The new functionality allows users to lock individual conversations, moving them into a dedicated section that is hidden from standard search results and chat lists. Once locked, a conversation is isolated from regular history, pinned conversations, and any default organization settings. This ensures that sensitive dialogues do not appear alongside routine queries or casual prompts.

Users can secure their locked chats with a four-digit PIN, or use Face ID or Touch ID on supported devices.

Unlike Temporary Chats, which create ephemeral sessions that vanish from history entirely, Locked Chats can be accessed at any time by the owner. This distinction provides a middle ground for users who need permanent but private records without exposing them to anyone with physical access to the device. The locked conversations remain stored on the user’s account but are shielded behind the chosen authentication method.

The feature is currently rolling out across a wide range of subscription tiers. It is available to users on ChatGPT Free, Go, Plus, Pro, Business, Enterprise, and Edu plans. However, the company has noted that the feature is not available on ChatGPT for Healthcare, likely due to existing regulatory and compliance frameworks surrounding medical data.

Privacy Implications and Limitations

While the locking mechanism provides a convenient barrier against visual eavesdropping, OpenAI has been transparent about its technical limitations. The company explicitly states that locking a chat does not create end-to-end encryption. Nor does it establish a separate encrypted vault for the data. This distinction is crucial for users who may conflate interface privacy with cryptographic security.

The feature does not alter how long OpenAI retains chat data, nor does it change the "Improve the model for everyone" setting. Furthermore, if the system had previously retained information from a conversation before it was locked, that memory persists. Locking the chat afterward does not purge previously remembered details or retroactively change data handling policies.

Additionally, the Locked Chats option is unavailable for several specific chat types. Users cannot lock temporary chats, archived chats, group chats, or conversations created from shared links. It is also incompatible with chats containing scheduled tasks or meeting recordings. These restrictions help maintain the integrity of features designed for collaboration and automation, which might be compromised by strict access controls.

Memory and Data Isolation

For users who have the Memory feature enabled, Locked Chats operates with a distinct, separate memory bank. This isolation ensures that insights gathered within a locked conversation do not bleed into the user’s general memory profile. New messages sent after locking a chat can update the locked memory but will not influence the regular memory used in other conversations.

This architectural choice prevents sensitive information from accidentally leaking into unrelated chats. It allows users to discuss personal matters or plan surprises without those details influencing the AI’s behavior in their daily work or casual interactions. The separation is automatic and enforced by the system once a chat is designated as locked.

Resetting the PIN associated with locked chats has severe consequences. OpenAI warns that resetting a PIN will permanently delete all locked chats, files, and other content protected by that previous PIN. This permanent deletion is irreversible, emphasizing the importance of securing authentication credentials. Users should treat their PINs with the same level of care they would apply to other sensitive digital keys.

Contextual Usage and Security Landscape

The release of Locked Chats comes as ChatGPT evolves from a purely productivity tool into a platform for personal reflection and private planning. As users increasingly confide in AI assistants, the need for granular privacy controls has become apparent. The feature is specifically tailored for situations where a user might be discussing something personal or planning a surprise.

It anticipates the common scenario of handing a phone to a friend or family member to show them something in a different chat. Without this feature, a glance at the recent history could inadvertently reveal private matters. By segregating sensitive topics, OpenAI aims to reduce social friction and anxiety around using AI in shared physical spaces.

However, security experts often caution that interface-level privacy is not equivalent to comprehensive data security. While Locked Chats limits access to the user interface, it does not address potential vulnerabilities in data storage or transmission. The feature serves as a practical solution for immediate, on-device privacy but does not solve broader concerns about how sensitive information is processed by the underlying models.

As the technology continues to mature, users are encouraged to evaluate their own risk tolerance. For low-stakes confidentiality, such as planning a birthday gift, the lock feature may suffice. For highly sensitive personal data, understanding the limits of the encryption and retention policies remains essential. The rollout began on Tuesday, and availability may vary by region and device as OpenAI expands access to the new capability.