Journalists from Android Authority, while analyzing a fresh build of the ChatGPT mobile app for Android (version 1.2026.230), found text strings in its code that describe in detail how a new feature — protected conversations — works. According to the descriptions found, users will be able to lock any saved chat, after which its name and contents will no longer appear in the regular sidebar, in the conversation history, or in the in-service search results. Returning to such a conversation will only be possible after entering a PIN code or verifying a fingerprint.

How conversation protection will work

According to the strings found in the code, the protection mechanism works on a 'one-time unlock' principle: a successful authentication via PIN code or biometrics only opens access to a specific locked chat, but does not remove its protection. The conversation will remain in a locked state until the user manually switches it back to normal mode. Thus, even when the device is unlocked, the chat will not be automatically 'unpacked' for viewing — each access to the contents requires a separate confirmation of the owner's identity.

Difference from already available temporary chats

Android Authority emphasize that the innovation should not be confused with the already existing temporary chats. The latter, by their nature, are not saved in the history: after the conversation ends, it is impossible to return to them, and the entire correspondence is erased. Locked conversations, on the contrary, will be fully stored in the user's account and will remain available to their owner at any time — provided a PIN code or fingerprint is entered. Judging by the description in the code, any saved chat can be protected this way, not only a specially created 'secret' one.

Data isolation and memory handling

The developers have also provided a special approach for data processing. Locked chats will be able to use information from ChatGPT's shared memory, however, new information that the neural network saves during a protected conversation will not be transferred to regular dialogs and will remain 'locked' inside the separate protected chats. In other words, information from such conversations will not automatically enter the model's shared memory, which reduces the risk of leaking confidential context into everyday conversations with the assistant.

Feature status and launch timeline

Despite the fact that the new build already contains fairly detailed descriptions of the upcoming feature, OpenAI has not yet announced its official launch. Because of this, it remains unknown whether conversation protection will be available to all users and exactly when it may appear in the release version of the app. At the moment, the information is based exclusively on the results of code analysis and is not confirmed by a press release or a statement from the developer.

Contradictory data

No direct contradictions in dates or figures between sources have been identified: both publications refer to the same Android build version (1.2026.230) and to the same source of discovery — Android Authority. The only caveat concerns the status of the feature: its presence is technically confirmed by code analysis, however, the fact and timing of the public launch have not been confirmed by OpenAI. Therefore, it is correct to speak not of a finished product, but of descriptions of a future capability found in the code, which may be released, modified, or canceled before the release.