Hand your phone to your partner for two seconds and watch your heart rate double. Not because you're doing anything wrong — but because your ChatGPT sidebar is sitting right there, full of conversations you'd rather keep to yourself. Maybe you asked it for advice about a medical scare at 2 AM. Maybe you were drafting a resignation letter. Maybe you were planning a surprise trip for someone who's literally in the car next to you. OpenAI's answer to that exact anxiety is the ChatGPT Locked Chats privacy feature 2026 — and it's the most useful privacy upgrade ChatGPT has shipped in a long time.
That exact, very human moment is what OpenAI finally decided to do something about. On Tuesday, October 20, 2026, the company announced Locked Chats for ChatGPT — a new privacy control that tucks selected conversations behind a four-digit PIN or your device's biometric unlock. It's rolling out gradually to every plan tier, from Free all the way up to Enterprise and Edu, and it's the first time a major AI chatbot has shipped anything like it. The announcement came at a moment when the ChatGPT platform has grown into one of the most-used digital tools on the planet, with hundreds of millions of weekly users sharing everything from code to heartbreaks with it.
But — and this is the part most headlines gloss over — Locked Chats is not encryption. It doesn't change what data OpenAI stores, and it won't erase anything the AI already remembers about you. The chatgpt locked chats privacy feature 2026 solves a real, everyday problem (nosy eyes on your screen), not the deeper question of who sees your data in the cloud. So what does it actually do? How do you use it? And is it enough? Let's get into all of it.
What the ChatGPT Locked Chats Privacy Feature 2026 Actually Is
Think of Locked Chats as a locked drawer inside your desk — not a bank vault in Switzerland. When you lock a conversation in ChatGPT, three things happen to it:
First, it disappears. The chat leaves your regular sidebar history, your pinned conversations list, and even the global search results inside ChatGPT. Hand your unlocked phone to someone, let them scroll your sidebar, and they will see absolutely no trace of that thread. Not the title. Not a preview. Nothing.
Second, it moves. Locked conversations live in a dedicated section under Projects > Locked Chats in the sidebar. The sidebar can still show the "Locked Chats" heading itself, but the titles inside stay hidden until you authenticate. On the web and in the desktop apps, that means entering the four-digit account PIN you set up. On iPhones and Android devices, you can also unlock with Face ID, Touch ID, or your device's own unlock method — fingerprint sensors on Android included — configured separately on each device as a convenience layer on top of the PIN.
Third — and this is genuinely clever — locked chats get their own isolated memory. With ChatGPT's Memory feature switched on, anything you mention inside a locked chat goes into a separate memory pool shared only with other locked chats. That detail about your divorce plans or your side hustle revenue won't suddenly surface in an unlocked work chat three days later. New messages sent after locking update only that isolated memory. It's a thoughtful design choice, and honestly one of the most under-appreciated parts of this update.
So who is this for? OpenAI's own examples are telling: discussing something personal, or planning a surprise, then handing your phone to a friend or family member to show them a different chat. The use cases write themselves. Couples who share tablets. Families where kids borrow mom's laptop for homework. Professionals who present their screen in Zoom meetings and don't want a thread titled "how to negotiate a raise with my boss" popping up mid-call. Freelancers who leave their laptop open at coffee shops. The chatgpt locked chats privacy feature 2026 is, at its core, a physical privacy tool — protection against the people who are already within arm's reach of your device, not against hackers, governments, or OpenAI itself.
And it's free. Not gated behind the $20-a-month Plus tier or the eye-watering $200-a-month Pro plan. OpenAI rolled it out to Free, Go, Plus, Pro, Business, Enterprise, and Edu users in stages. The one notable exclusion is ChatGPT for Healthcare, which makes sense — healthcare deployments run under stricter compliance regimes where a PIN folder doesn't map to any regulatory framework. If you don't see "Lock chat" in your menu yet, just update the app and wait; it's a phased rollout.
How to Lock (and Unlock) a ChatGPT Conversation: Step by Step
Setup takes about thirty seconds. OpenAI kept the flow dead simple, which matters — privacy features nobody can figure out might as well not exist.
Locking an existing conversation
Open the conversation you want to hide and tap the three-dot menu next to its title. You'll see a new option: "Lock chat." Tap it, confirm, and if this is your first time, ChatGPT will ask you to create a four-digit account PIN — you type it twice, just like setting a passcode on your phone. That's it. The conversation immediately vanishes from your regular history and relocates to Projects > Locked Chats. Done. It's the same menu flow whether you're on the web, desktop, or mobile.
Starting a brand-new locked chat
You don't have to lock something old — you can also begin a conversation directly inside the protected space. Go to Projects > Locked Chats, authenticate with your PIN or biometrics, and type your message. The conversation is saved inside the locked section from birth and never touches your regular history or search. This is the move if you know upfront that the thread is sensitive — say, before typing that very personal medical question.
Unlocking and using a locked chat
Head to Projects > Locked Chats, enter your PIN (or glance at your phone for Face ID), and your protected conversations appear. Chat inside them normally. When you're done, you don't have to do anything: on the web and desktop, locked chats relock automatically when you leave the tab or window, refresh the page, or close the tab. On iPhone and Android, access ends after about one minute away from a locked chat, or one minute with the app in the background. There's also a "Lock now" button if you want instant protection — like when someone walks into the room mid-conversation. Signing out or switching accounts locks everything too.
The one thing you must know about your PIN
Your account PIN is tied to your OpenAI account, not to a specific workspace — so if you use multiple accounts, manage each PIN separately. And here's the critical bit: if you forget your PIN and reset it, every locked chat, every file, and every piece of memory protected by the old PIN is permanently deleted. No recovery. No "forgot PIN" lifeline. OpenAI is explicit about this, and it's clearly deliberate — it means they can't be compelled to hand over your locked chats, because they have no way to restore them either. But it also means a forgotten PIN equals total data loss. Write the PIN down somewhere safe, away from your devices. Seriously. A sticky note in a drawer beats a deadlocked drawer you can never open.
What Locked Chats Does NOT Do: The Honest Limits
Here's the thing. A feature that looks like security but gets confused for privacy can be dangerous in its own way — it makes people over-trust it. So let's be brutally honest about what the chatgpt locked chats privacy feature 2026 does not do, straight from OpenAI's own help pages and support documentation.
It does not encrypt your conversations end-to-end. This is the big one. A locked chat is hidden from your screen — it is not scrambled into a vault that only you can read. OpenAI can still access locked conversations. Nothing stops your data from traveling to and living on OpenAI's servers, same as before.
It does not change data retention. Your locked chats are stored for as long as OpenAI's normal retention policy says, which is the same as unlocked chats. Locking a chat doesn't set it to expire faster, and it isn't the same as deleting it. If you want a conversation gone from OpenAI's systems, you have to delete it — the lock is not a delete.
It does not turn off training-data use. The "Improve the model for everyone" setting — the toggle that lets OpenAI use your conversations to train future models — is completely unaffected by locking a chat. If you have it enabled, your locked chats can still be used for training. If it's off, they're excluded. The lock changes nothing either way, which surprises a lot of people. Hide the chat from your sister: yes. Hide it from the model trainers: no.
It does not erase what ChatGPT already remembered. Locking a conversation only partitions memory going forward. Anything the AI already saved into your regular Memory before the lock stays right there in your regular Memory. So if you chatted for a month about your health problems in an unlocked thread and then lock it today, the old health details are still sitting in your unlocked memory pool. The isolation starts at the moment of locking, not retroactively. If this worries you, go review your saved memories under Settings and delete the ones you don't want — the lock won't do it for you.
It does not override workplace or legal rules. OpenAI is explicit: the lock doesn't override applicable workspace policies, safety requirements, or legal and compliance obligations. If you're on a company Business or Enterprise account, your admin's data controls still apply. And obviously, a PIN doesn't protect you from a lawful data request or a determined forensic analysis of your device.
It can't lock everything. Temporary chats can't be locked (they don't persist anyway). Neither can archived chats, group chats, chats opened from a shared link, chats with scheduled tasks, or chats containing meeting recordings. So if you were hoping to lock your work meeting transcripts — nope, not supported.
Does all this make Locked Chats useless? Not at all. But it reframes what it is: a tool against the people near you, not against the systems behind you. Which is fine — the people near you are the actual threat in most real lives. OpenAI, for its part, has had a busy privacy season beyond this feature; the company's internal culture has been in the spotlight too, as we covered in our piece on OpenAI firing three researchers over policy disagreements. The throughline is that AI privacy is getting more complicated, not simpler.
Locked Chats vs. WhatsApp Chat Lock, Signal, and Google's Locked Folder
OpenAI didn't invent the "hide this chat" concept. Messaging apps have been doing versions of this for years. But the comparison is actually fascinating, because the differences reveal a lot about what AI privacy is — and isn't.
WhatsApp Chat Lock
Meta's WhatsApp introduced chat lock back in 2023: you lock individual chats, they move into a "Locked chats" folder, and you unlock with your device PIN, Face ID, or fingerprint. Sound familiar? It's basically the exact same mental model — physical privacy against whoever picks up your phone. The key difference: WhatsApp pairs its chat lock with end-to-end encryption by default, so locked or not, Meta can't read your messages. ChatGPT has no end-to-end encryption at all, locked or unlocked. So the lock mechanics are twins, but the underlying privacy guarantees are very different animals.
Signal's disappearing messages
Signal takes the opposite philosophy. Instead of hiding conversations, it lets them destroy themselves — disappearing messages that vanish after a timer you set, with true end-to-end encryption under the hood. It's the nuclear option: nothing to find, because there's nothing left. ChatGPT Locked Chats is the preservation option: the conversation is saved and searchable by you forever, just hidden from view. For AI chats, which you often want to revisit for reference — that long debugging session, that business plan draft — the lock model arguably fits better than disappearing messages. But Signal wins on actual cryptographic privacy by a mile.
Google's Locked Folder (Photos)
Google Photos' Locked Folder is probably the closest sibling. It hides selected photos and videos behind a device passcode, and crucially, items in the Locked Folder don't sync to the cloud and don't appear in the main grid, memories, or search. It's local-first privacy. ChatGPT's locked chats, by contrast, still live in your cloud account and sync across devices — that's why your one account PIN works everywhere. More convenient, less private. The trade-off is real, and it tells you everything about how AI companies think about data: your chats are a cloud product first, a privacy surface second.
The honest scorecard
If your threat model is "my roommate borrows my laptop," Locked Chats is genuinely great — it's as good as WhatsApp's lock for that purpose. If your threat model is "I don't want a corporation training models on my therapy sessions," the lock does nothing — go look at your training-data toggle instead. And if your threat model is "I need nobody on earth to read this," you need Signal, not a chatbot. Knowing which one you are is the whole game.
ChatGPT Privacy Settings to Use Alongside Locked Chats
Locked Chats is one piece of the puzzle. If you actually care about your AI privacy — and by 2026, you should — here's the full toolkit, because a locked drawer in an otherwise open house still leaves you exposed.
1. Check "Improve the model for everyone." This is the single most important privacy toggle in ChatGPT, and most people have never looked at it. In Settings > Data Controls, you'll find the switch that decides whether OpenAI can use your conversations to train future models. On free-tier accounts, this has historically defaulted to on. Turn it off if you're uncomfortable with human reviewers or training pipelines ever seeing your chats. Remember: the lock doesn't change this. It stays exactly however you set it.
2. Audit your Memory. ChatGPT's Memory feature quietly saves facts about you — your name, your job, your health concerns, your preferences. Some of that is wonderfully convenient. Some of it you'll wish you'd never typed. Go to Settings > Personalization > Memory and actually read what's in there. Delete anything sensitive. And if you'd rather the AI remember nothing at all, you can switch Memory off entirely.
3. Use Temporary Chats for the truly sensitive stuff. OpenAI's Temporary Chats — the feature that existed before Locked Chats — are sessions that never save to your history and aren't used for training or memory. If you're asking about something you'd never want stored anywhere, Temporary Chat is the right tool, not a locked chat. Locked chats are for conversations you want to keep but hide. Temporary chats are for conversations you want to have and forget. Learn the difference and use both.
4. Review your chat history regularly. Old conversations accumulate. Every few months, scroll through your history and archive or delete threads you no longer need. Deleted chats move to a trash state and are permanently removed after 30 days. It's digital spring cleaning, and it shrinks your exposure surface for free.
5. Be careful what you paste. This one's on you, not OpenAI. People paste Social Security numbers, bank statements, medical records, and unreleased business plans into chatbots every single day. A locked chat doesn't un-paste your Social Security number from OpenAI's servers. As a rule of thumb: if you wouldn't email it to a stranger, don't paste it into an AI chat — or at least strip the identifying details first.
6. On shared devices, use separate profiles. If your family shares one laptop, browser profiles are your friend. A locked chat protects against casual scrolling, but a separate browser profile with its own login protects against everything short of targeted snooping. Layer your defenses.
One more thing worth noting: privacy in AI products is now moving fast across the whole industry. OpenAI's push into privacy controls comes the same month as big moves in AI hardware and infrastructure — we recently wrote about Apple's touchscreen MacBook event and Amazon's new AI training chips, and the thread connecting all three stories is the same: AI is touching more of your life, so the question of who sees that data is getting louder every week.
What the ChatGPT Locked Chats Privacy Feature 2026 Signals About AI Privacy
Step back from the how-to for a second, because there's a bigger story here. When the world's most-used AI chatbot ships a feature designed around people looking over your shoulder, it's an admission of something important: AI has become intimate.
We don't lock things we don't care about. Nobody puts a PIN on their weather app. The fact that millions of users asked — implicitly, through their behavior — for a way to hide AI conversations means those conversations now contain the kind of material people historically kept in diaries, therapists' offices, and lawyers' filing cabinets. That's a profound shift in how humanity relates to software, and it happened in about four years.
It also signals where AI privacy is headed in 2026 and beyond. Notice what OpenAI didn't ship: end-to-end encryption for chats. No zero-knowledge architecture. No local-only processing mode for sensitive threads. The industry is giving users interface-level privacy — controls over what appears on screen — while keeping the cloud-based data model intact underneath. That's not an accident. AI models need your data to improve, and the business model of the biggest chatbot on earth depends on the flywheel of usage, data, and better models. Real cryptographic privacy would break that flywheel. So we get locks instead of encryption, and the gap between the two is where the next decade of AI privacy debates will live.
You'll also start seeing this pattern everywhere. WhatsApp-style chat locks are already migrating from messaging apps into AI products, and it's a safe bet that Google's Gemini, Anthropic's Claude, and Microsoft's Copilot are watching this rollout closely. Within a year, "lockable AI chats" will probably be table stakes across the industry — just like dark mode or read receipts were in messaging. The feature that feels novel today will be the feature you'd be shocked to find missing tomorrow.
And here's my honest take as someone who writes about tech for a living: Locked Chats is a genuinely useful, well-designed feature that solves the problem it claims to solve — and no more. The danger is only if you mistake it for something it isn't. A hidden chat is not a private chat in the cryptographic sense. OpenAI has been admirably transparent about the limits (which is more than some companies would do), but transparency in a help-center article doesn't reach the average user who just taps "Lock chat" and assumes the padlock icon means padlock-grade security. If this article does one thing, let it be this: teach the people you share devices with the difference between hidden from view and protected from access. They are not the same thing.
Frequently Asked Questions About Locked Chats
Is ChatGPT Locked Chats free?
Yes. The feature is rolling out to every plan tier — Free, Go, Plus, Pro, Business, Enterprise, and Edu — at no extra cost. The only plan excluded is ChatGPT for Healthcare, which operates under separate compliance rules. You don't need the $20-a-month Plus subscription to use it.
Does locking a chat encrypt it?
No. Locked Chats adds no end-to-end encryption and creates no separate encrypted vault. It hides the conversation from your sidebar, pinned chats, and search results, and requires a PIN or biometric to open. OpenAI can still access the conversation on its servers, and your normal data-retention and model-training settings continue to apply.
What happens if I forget my Locked Chats PIN?
If you reset a forgotten PIN, OpenAI permanently deletes all chats, files, and locked-chat memory tied to the old PIN — with no way to recover them. If you remember your current PIN, you can change it without losing anything. Bottom line: store your PIN somewhere safe and separate from your devices.
Can I lock a chat on one device and unlock it on another?
Yes. Your four-digit account PIN is tied to your OpenAI account, so it works on every device where you're signed in. Biometric unlock (Face ID, Touch ID, Android fingerprint) is configured per device as an optional convenience on top of the PIN. Access relocks automatically — on web when you leave the tab, on mobile after about a minute of inactivity — or instantly with the "Lock now" button.
Do locked chats still train OpenAI's models?
Locking a chat does not change your "Improve the model for everyone" setting. If that toggle is on, locked chats can still be used for training; if it's off, they're excluded. The lock itself neither enables nor disables training-data use. Check Settings > Data Controls to see your current setting — it's the toggle that actually matters for this question.
The Bottom Line
The chatgpt locked chats privacy feature 2026 is one of those updates that's easy to underestimate and easy to overestimate at the same time. Underestimate it, and you miss a genuinely practical tool — the next time your kid grabs your phone, your colleague glances at your screen share, or your partner borrows your laptop, those locked conversations stay exactly where they belong: out of sight. Overestimate it, and you'll sleepwalk into believing a screen-level lock equals real data privacy, while your chats still sit on servers, still feed training pipelines if you've opted in, and still carry everything the AI remembered about you before you ever tapped "Lock chat."
The smart move? Use it. Lock the conversations that deserve a locked drawer. But pair it with the real controls: check your training-data toggle, audit your Memory, use Temporary Chats for the things you want forgotten, and never paste anything into any AI chatbot — locked or not — that you couldn't live with existing on a server somewhere. A PIN keeps prying eyes out of your phone. Good habits keep your data out of the wrong hands entirely.
AI is only getting more personal from here. Locked Chats is the industry's first acknowledgment of that intimacy — a small padlock icon that says, "we know you're telling this thing your secrets now." The next step, whenever it comes, is giving users privacy that's real under the hood, not just on the screen. Until then, lock your chats, mind your settings, and don't confuse hidden with safe.