ChatGPT adds 4-digit PIN locked chats for Free, Go, Plus and Pro users, but without end-to-end encryption

OpenAI is rolling out locked chats in ChatGPT, a feature that hides selected conversations behind a 4-digit PIN. The company says the rollout is gradual and covers Free, Go, Plus and Pro accounts. Engadget reported the launch on October 20, 2026.

How locked chats work

When a chat is locked, it no longer appears in regular chat history, pinned chats or conversation search. Locked conversations are grouped in a new project called "Locked chats." Opening that project requires the account's 4-digit PIN.

On iPhone and Android, users can unlock with biometric authentication such as Face ID instead of typing the PIN. On Android and iOS, voice mode works together with locked chats.

What locked chats do not do

OpenAI states the limits explicitly on a newly published support page:

• Not end-to-end encryption. Locked chats are an access barrier within the app, not a change to how conversations are encrypted.
• No legal shield. The feature does not alter OpenAI's legal obligations, and it does not protect users from workplace or legal consequences.
• No change to training settings. Locking a chat does not automatically change a user's model training preferences.
• No change to retention. It does not alter how long OpenAI retains the conversations.
• No Codex support. Codex conversations cannot be locked.

PIN reset deletes locked content

Resetting the account PIN deletes all locked chats and any files uploaded to the Locked chats project space. OpenAI warns: "This cannot be undone, and OpenAI Support cannot recover the deleted content." Users who forget their PIN lose that data permanently.

Availability

The rollout is gradual. Users who don't see the option immediately may need to wait a few days. Free, Go, Plus and Pro accounts are included. The source material does not say whether Business, Enterprise or Education plans are covered.

This is a product feature, not a model release. OpenAI has not announced any new model, context window, pricing or benchmark changes alongside it.

What this means

Locked chats address a narrow threat model: someone glancing at your screen or borrowing your unlocked phone. They do not address data handling on OpenAI's side. Because retention and training settings are unchanged, a locked chat is stored and processed like any other. Users with sensitive material should not treat the PIN as a privacy guarantee against the provider, legal process or an employer's policies.

The PIN-reset behavior is the main operational risk. Because OpenAI cannot recover deleted locked content, the feature trades convenience for permanence. Anyone storing important conversations or files in the Locked chats project should keep copies elsewhere.

The exclusion of Codex also leaves a gap for developers who use ChatGPT for coding work involving credentials or proprietary code. For that work, locked chats are not a substitute for proper secrets management or an enterprise data-controls plan.