ChatGPT is adding a privacy-control feature called Locked Chats, giving users a separate place for conversations they do not want casually visible in the app’s normal history. The feature is intended to put a practical barrier between private exchanges and anyone who happens to be looking at an unlocked phone or shared screen—not to transform those conversations into encrypted vaults beyond the service’s reach.
With Locked Chats enabled, users can choose whether a previous conversation is kept out of regular chat history, pinned chats and conversation search. Locked conversations are gathered inside a new project labeled “Locked chats”. Opening that area requires a four-digit PIN. On iPhone and Android, users can also authenticate with the device’s supported systems, including Face ID.
That distinction matters. A person handing over their phone to show a friend a game clip, pass it to a family member, or let someone make a quick call may not want that person to immediately see a sensitive ChatGPT thread in the sidebar. Locked Chats is built for that everyday, local privacy problem: reducing the chance of casual viewing by somebody who has temporary access to a device.
What Locked Chats changes
The visible effect is straightforward. A chat that has been locked can be excluded from the three usual places where an older exchange could otherwise be easy to stumble across:
- the standard chat-history list;
- pinned conversations; and
- conversation search.
Instead, those exchanges live in the dedicated Locked Chats project. Requiring a PIN or phone-based authentication before displaying that project makes the feature useful against opportunistic access. It also gives users a clearer way to separate ordinary prompts from conversations they would rather not leave in plain sight.
For people using ChatGPT around games, that could mean keeping personal planning conversations separate from everyday questions, or preventing a visitor with access to an unlocked device from browsing through old discussions. The value is not tied to one kind of prompt. It is the ability to prevent a past conversation from appearing automatically in the familiar places where ChatGPT normally surfaces it.
The rollout is gradual, so the option may not be present for every account immediately. The feature is planned for free users and subscribers on Go, Plus and Pro. If it has not appeared in settings or available chat controls yet, the rollout may take several days to reach a particular account.
A lock screen for chats, not encryption
The essential limitation is explicit: Locked Chats is not end-to-end encryption.
End-to-end encryption is a technical design in which content is encrypted so that only the communicating endpoints are intended to read it. A PIN gate inside an app serves a different purpose. It controls access to a section of the interface after someone is already using the relevant device or account. It can be valuable, but it is not a statement that the service cannot access the underlying information.
Users should therefore avoid treating a locked chat as a method for bypassing legal obligations, workplace rules or other external requirements. The feature is not positioned that way. Its role is narrower and more practical: keeping conversations out of view when someone is nearby or when a phone is loaned to another person.
This is an important expectation-setting point because terms such as “locked,” “private” and “hidden” can sound stronger than the protections they actually describe. Here, the benefit is primarily visibility control. A four-digit PIN and biometric check can stop a passerby or a casual borrower from opening the project, but neither label alone should be read as a guarantee of end-to-end encrypted storage or a change in the platform’s obligations.
Training and retention settings do not change automatically
Putting a conversation into Locked Chats does not automatically alter a user’s model-training settings. It also does not automatically change how long the service retains those conversations. Those are separate choices and policies from the interface-level lock.
In practical terms, users who care about both casual device privacy and how their data is handled need to understand that these are two different questions:
- Who can see a conversation on my phone right now? Locked Chats can help manage that.
- What settings apply to the conversation beyond my device? Locking it does not itself change training or retention settings.
That separation is useful because it prevents a common security mistake: assuming one safeguard automatically supplies every other safeguard. A private folder can be a meaningful extra layer for on-device viewing while leaving account-level and service-level settings untouched.
The same care applies when using AI tools alongside gaming communities and game-related work. An AI-assisted project can involve drafts, technical notes or other material worth keeping away from casual eyes, but a local lock should not be confused with a full data-governance setting. For another example of AI entering a gaming-adjacent workflow, see this report on a fan-made Oblivion Remastered PC patch with AI-assisted work.
PIN resets carry a permanent deletion risk
The sharpest practical warning attached to Locked Chats concerns the PIN reset process. Resetting the account PIN deletes every locked chat, as well as files uploaded to that project space. The deletion cannot be undone, and support cannot recover the removed material.
That makes the PIN more consequential than an ordinary convenience setting. Users considering Locked Chats should choose a PIN they can reliably remember and should understand the cost of losing it. If a conversation or uploaded file matters, the feature’s reset rule means it is unwise to assume recovery will be available later.
This is not merely a minor inconvenience. A locked project may contain several conversations and associated uploads, and the stated consequence applies to all of that content when the PIN is reset. The trade-off is clear: a stronger barrier against someone casually opening the area is paired with an irreversible loss path if the user must reset the credential.
Before moving a chat into the locked project
- Decide whether the concern is casual on-device access, since that is the problem the feature is designed to address.
- Remember that locking does not automatically adjust model-training or retention settings.
- Choose a four-digit PIN with the awareness that a reset deletes locked chats and project uploads.
- Do not rely on Locked Chats as end-to-end encryption or as a way around workplace or legal obligations.
- Check whether the feature has reached the account, as availability is rolling out over time.
Mobile support and the Codex exception
On Android and iOS, Locked Chats can be used alongside voice mode. That means the addition is not limited to text-only use on phones. The mobile biometric option also provides an alternative to entering the four-digit PIN when supported authentication is available on the device.
There is one stated exception: Codex conversations cannot be locked with this feature. Users who rely on Codex should not expect those chats to be moved behind the Locked Chats gate. The limitation is particularly worth noting for anyone who treats all ChatGPT conversation types as interchangeable; this privacy control does not apply across every conversation category.
Locked Chats is therefore best understood as a targeted organizational and privacy layer. It can keep selected standard conversations out of common browsing surfaces, group them in one dedicated project, and require a PIN or supported biometric check to open them. It does not change the fundamental nature of the service’s encryption, retention or training controls, and it comes with a strict deletion consequence if the PIN must be reset.
For users whose concern is someone glancing at an unlocked screen, that narrow focus may be precisely the point. For users seeking stronger guarantees about data handling, the feature’s stated limits are just as important as its new lock icon.
