Like WhatsApp, ChatGPT finally puts your conversations under lock and key.

Since October 20, OpenAI has been rolling out “locked chats” in ChatGPT, from free accounts to Enterprise and Edu plans. The feature is reminiscent of the chat locking that WhatsApp has offered since May 15, 2023: a four-digit code, or Face ID and Touch ID on mobile, and the thread disappears from the screens.The feature, identified as early as August in the code for the Android app, is gradually rolling out to the web, phones, and desktop apps.

A passcode-protected folder, hidden from history

Once locked, the conversation moves into a dedicated space and is removed from history, pinned threads, and search results. Unlocking it requires the code or the phone’s biometric authentication, and a single gesture is then enough to return the thread to the regular flow. The “Locked Chats” folder remains visible in the interface. WhatsApp has offered a secret code since late 2023 that conceals even the folder’s existence—a level of discretion still beyond ChatGPT. As the rollout stretches over several weeks, the option may still be unavailable on some accounts.

Temporary chats, archives, group discussions, and conversations linked to a scheduled task remain out of scope. On computers, the latest update to the macOS and Windows apps is required, with a gradual rollout still ongoing according to OpenAI. Memory is partitioned in the same way. Locked threads can draw on the assistant’s general memories, but the reverse never happens: what is discussed behind the code feeds a separate memory, which does not color your ordinary conversations. Two notebooks, only one of which is left lying on the living-room table.

No encryption, and a passcode that can take everything with it

OpenAI presents the feature as a privacy screen, designed for a phone being passed from hand to hand or for prying eyes on the subway. The conversations’ retention period remains unchanged, as does the setting that allows models to be trained on your exchanges, and end-to-end encryption remains absent. The lock offers no greater protection against a judicial order either. In companies and universities, locked threads remain recorded in the compliance logs accessible to the employer.

Resetting the passcode permanently deletes the locked conversations, their memories and the files they protected. OpenAI’s help page warns bluntly: “This action is irreversible, and support cannot recover deleted content”. Administrators of a professional workspace can neither reset an employee’s passcode nor disable the feature. Forgetting your own, however, destroys nothing as long as the reset is not initiated.

According to the Arcep Digital Barometer published in February, 48% of French people aged 12 and over use generative AI, a fifteen-percentage-point increase in one year. And 63% of these users go to ChatGPT first.The medical, financial or romantic questions that used to end up in a search engine now accumulate in discussion threads that remember everything, hence the appeal of drawing a curtain.

👉🏻 Follow the latest tech news in real time: add 01net to your sources on Google and subscribe to our WhatsApp channel.

Source: OpenAI