To further protect users’ privacy and security, OpenAI has recently officially launched a brand-new ChatGPT feature — “Locked chats” — to users worldwide.
As for how it works in practice: when a user chooses to enable the feature for a specific sensitive conversation, that conversation is automatically removed from the regular chat history, the pinned list, and ordinary search results, and is moved in its entirety into a separate “Locked chats area.” When the user later wants to reopen these locked conversations, they must enter the four-digit account PIN for verification; and on mobile devices with the corresponding hardware, users can also complete identity verification directly through biometrics such as Face ID or fingerprint, or through the device unlock method.
At present, the feature is being gradually rolled out to users of various paid and free plans, although the Healthcare version is not yet within the scope of support. The core purpose of this design is to effectively prevent others from directly viewing conversations involving personal privacy or sensitive content when a user lends their device to someone else.
In terms of data storage and product logic, OpenAI has drawn careful boundaries: locked conversations and their associated files remain fully saved in the user’s account, and this area has a memory space independent of ordinary conversations; memory information already generated before locking will not be deleted.
It should be noted in particular that “Locked chats” essentially only changes where conversations are displayed, their entry point, and the access verification method; it does not provide end-to-end encryption, nor does it change the existing chat retention period or model-training-related settings. The core positioning of this feature is to prevent others from directly viewing conversations, and it is not equivalent to deep encryption or a change to the underlying data processing methods.