A lock for your most private conversations
OpenAI has started rolling out a new privacy feature for ChatGPT called Locked Chats, announced on Tuesday, 20 October 2026. It lets you protect individual conversations with a four-digit PIN or a biometric unlock such as Face ID or Touch ID (LiveMint). The idea is simple: if you share a phone or computer, or someone can scroll through your history over your shoulder, your most sensitive conversations stay hidden.
Once locked, a conversation moves into a dedicated Locked Chats section and disappears from your regular history, pinned chats and standard search results (The Hindu). You can lock existing conversations, not just new ones (The Hindu), and opening one requires your PIN or biometric unlock (LiveMint). On Android, screenshots of locked chats may appear blank or black (The Hindu).
Locked chats are not deleted chats — unlike Temporary Chats, they stay in your account (LiveMint). OpenAI positions the feature for personal matters and for planning things like a surprise trip (LiveMint).
The catch: a lock is not a vault
Here is the part worth reading twice. OpenAI states that locked chats have no end-to-end encryption and are not a locked vault in any cryptographic sense (LiveMint). Locking a chat does not change OpenAI’s data retention policy, does not delete anything ChatGPT remembered before you locked it, and does not change your “Improve the model for everyone” training setting (LiveMint). Locked chats may even still show ads (The Hindu).
In plain terms: the PIN protects your chats from people with physical access to your device. It does not change how OpenAI itself stores or handles the conversation.
The warning that actually matters: forget the PIN, lose everything
If you reset your account PIN, all of your locked chats — conversations and uploaded files alike — are permanently deleted, and OpenAI’s support team cannot restore them (LiveMint). The Hindu confirms the same: resetting the PIN permanently deletes all locked chats and the content tied to the old PIN.
So the one thing you must not do is treat that PIN casually. A forgotten PIN does not mean a quick reset and carry on — it means the locked archive is gone for good.
One more quirk: locked chats remember separately
When Memory is enabled, locked chats build their own separate memory. New messages in a locked chat can add to that separate memory, but they do not update your regular ChatGPT memory (LiveMint). It is a sensible privacy boundary — locked conversations will not leak context into your everyday chats — but it does mean ChatGPT may “know” different things about you in the two spaces.
Who gets it — and what can’t be locked
The feature is available across OpenAI’s Free, Go, Plus, Pro, Business, Enterprise and Edu offerings, with the rollout still ongoing (The Hindu). It is not available on ChatGPT for Healthcare (LiveMint).
Some things cannot be locked at all: Temporary Chats, archived conversations, group chats, chats from shared links, chats with scheduled tasks and chats containing meeting recordings (LiveMint). And because the rollout is gradual, the option may not have reached your account yet.
What this means
Locked Chats is a genuinely useful feature — the digital equivalent of a drawer with a lock on it, for the conversations you would rather your family, flatmates or colleagues never stumble across. We have written before about the privacy questions that follow every new AI product, from always-listening wearables to AI agents acting on your behalf.
But manage your expectations. This is access protection, not data protection: it keeps a curious family member or colleague out of your chats, while leaving OpenAI’s retention and training settings exactly as they were. And the permanent-deletion-on-reset rule turns a humble 4-digit PIN into something closer to a master key.
Use it for what it’s good at — privacy from the people around your screen. Just write the PIN down somewhere safe first.