ChatGPT Locked Chats Keep Private Threads Behind a PIN

OpenAI has begun rolling out ChatGPT Locked Chats, a privacy feature that moves individual conversations out of the regular chat history and behind a four-digit PIN or biometric unlock. The feature was announced this week and is reaching users on the Free, Go, Plus, Pro, Business, Enterprise and Edu plans in stages, according to Livemint's reporting on the launch. Once a conversation is locked, it leaves the main history, pinned chats and search results, and lives in a dedicated section under Projects until the owner unlocks it with a PIN or a supported biometric.

The feature is aimed at anyone who shares a phone or computer with other people. According to OpenAI, it suits conversations about personal matters or plans such as a surprise trip, the kind of threads users may not want visible when someone else picks up their device. The rollout covers seven subscription tiers, though notably not the Healthcare plan, a broad sweep that suggests OpenAI expects locked chats to become an everyday control rather than a niche extra.

Locking a conversation is deliberately simple. From the three-dot menu beside any chat, users pick Lock chat, confirm the choice, and set an account PIN by entering it twice, as Engadget's coverage describes. That same PIN then works across every device signed in to the same account. On iPhones and Android phones, users can instead unlock with Face ID, Touch ID or the device's own unlock method, which can be switched on separately on each device as a convenience on top of the PIN.

Users can also start a protected conversation from scratch. New chats can be created directly inside the Locked chats area under Projects, so a sensitive thread never has to sit in the open history at all. Files get the same treatment: anything uploaded or created inside a locked chat can only be opened from within that chat, and stays hidden from the broader file storage area. Files pulled in from outside the conversation keep their original access rules rather than inheriting the lock.

Your memory gets split in two

The feature also changes how ChatGPT handles context behind the scenes. Locked chats run on their own isolated memory partition. As OpenAI explains on its support page for the feature, new messages sent in locked chats will not update the memory used in regular conversations. In practice, personal details discussed inside a locked thread will not resurface later in everyday work chats. With ChatGPT Locked Chats, that separation is the point: a private conversation should not leak into future answers.

The split comes with an important caveat. Locking a chat does not erase what ChatGPT remembered before the lock was applied, so details already stored in the main profile stay there. And the memory wall works in one direction for context: a locked conversation can still draw on relevant information from regular Memory when that feature is enabled. Users who assume everything discussed before locking vanishes will be disappointed. The lock only walls off what comes after.

The locks are not encryption

OpenAI is explicit that ChatGPT Locked Chats is not an encrypted vault. According to the company's help documentation, locking a conversation does not apply end-to-end encryption, does not change how long the company keeps the chat, and does not alter the setting that lets conversations be used to improve its models. The feature protects against a different threat entirely: the casual onlooker, the shared device, the borrowed laptop. Against anyone who can compel or access OpenAI's servers, it offers nothing new.

The PIN itself is the real line of defense, and OpenAI built it to be unforgiving. If the PIN is forgotten and has to be reset, every locked chat and any file protected only by the old PIN is permanently deleted, and even OpenAI's support team cannot recover them. Workspace administrators cannot reset someone else's PIN either. Changing a known PIN, by contrast, deletes nothing. The rule is blunt: lose the code, lose the chats.

That design draws a clear line between Locked Chats and the existing Temporary Chats mode. Temporary chats never appear in history at all and vanish after the session, while locked conversations persist and can be revisited any time after authentication. For people who want privacy plus continuity, the new mode fills the gap. For people who want no record at all, temporary chats remain the answer.

There is also a counterpoint worth keeping in mind. Anyone who can open a locked chat can still copy its contents out, so the feature guards access on the device rather than controlling what leaves it. The keep-it-private instinct is showing up elsewhere in mobile AI too, from on-device scam warnings that never upload your messages to AI chat apps folding third-party tools into a single window. As assistants absorb more personal context, device-level locks like this are likely to become standard, but they are a lock on the door, not a safe. For now, that lock is rolling out to millions of users who share their screens with roommates, coworkers and curious friends.