ChatGPT may be preparing a new privacy feature that would allow users to place individual conversations behind an additional security check, potentially making it easier to keep sensitive AI interactions away from anyone who has access to an unlocked phone.
References to a “locked chats” feature have been discovered in code inside the latest Android version of ChatGPT. The currently inactive functionality appears designed to remove protected conversations from the app’s normal sidebar, chat history and search results. Access could reportedly require a PIN or biometric authentication such as a fingerprint.
There is no official confirmation from OpenAI that locked chats will be released, however. Features found in application code can be abandoned, substantially changed or remain under development for long periods. There is also no confirmed launch date or indication of whether the feature would arrive simultaneously on Android, iOS and the web.
Still, the idea addresses an increasingly relevant privacy problem for AI assistants. ChatGPT conversations can contain considerably more personal information than a typical search history. Users routinely turn to generative AI for help drafting workplace documents, discussing personal situations, planning purchases and exploring questions they may not want visible to somebody casually using their device.
Most smartphones already provide device-level protection through passcodes and biometric authentication, but that security becomes less useful once the device has been unlocked. An individual chat lock would provide another boundary inside ChatGPT itself, similar in principle to privacy features that messaging and photo applications have increasingly adopted.
The potential feature also appears distinct from ChatGPT’s existing Temporary Chat option. Temporary conversations are intended for interactions users do not want retained as part of their regular ChatGPT history. Locked chats, based on the code references uncovered so far, would instead remain available while being concealed from the app’s usual conversation surfaces.
That distinction raises an important unanswered question: memory. ChatGPT can use saved memories and, depending on settings and product behavior, information from previous conversations to make future responses more relevant. It is not yet clear whether information contained inside a locked conversation would remain accessible to those systems, be isolated from ordinary chats, or operate under separate memory controls.
That detail matters because hiding a conversation from the interface is not necessarily the same thing as isolating its information from other ChatGPT features. For users handling particularly sensitive material, the difference between visual privacy, account privacy and data retention can be significant.
If OpenAI does release locked chats, the feature would represent a fairly small interface change with broader implications for how people use AI assistants. As these tools become repositories for work, personal questions and long-running conversations, privacy controls increasingly need to operate at the conversation level rather than simply relying on the security of the device surrounding them.
