Is ChatGPT Private? What Happens to Your Personal Chats

Is ChatGPT private? Private from other users, yes, unless you share a link. Confidential, no. Your chats are stored by OpenAI until you delete them, can be used to train its models unless you opt out, have no therapist-patient privilege, and can be demanded in a lawsuit. If you vent to ChatGPT about your relationship, the more immediate risk may be simpler: someone picking up your phone.

Here's what actually happens to your chats, the settings worth changing, and what to keep out of any AI tool.

Is ChatGPT confidential?

No, and OpenAI's own CEO has said so. On a July 2025 episode of Theo Von's podcast, Sam Altman said people use ChatGPT "as a therapist, a life coach," for relationship problems, and that unlike with a therapist, lawyer or doctor, "we haven't figured that out yet for when you talk to ChatGPT." If there were a lawsuit, he said, OpenAI could be legally required to produce those conversations.

That isn't hypothetical. In its copyright lawsuit against OpenAI, The New York Times asked the court to make OpenAI preserve user chats, including deleted ones. For several months in 2025 OpenAI was under an order to retain consumer ChatGPT content. OpenAI says that obligation ended on September 26, 2025, but it is still holding a limited set of user data from April to September 2025 under legal hold. Separately, in late 2025 a federal judge ordered OpenAI to hand over roughly 20 million de-identified ChatGPT logs to the plaintiffs.

None of that means anyone is reading your Tuesday-night vent. It does mean "private" has limits you should know about.

Does ChatGPT save your conversations?

Yes. According to OpenAI's help center:

• Regular and archived chats stay saved to your account until you delete them. Archiving hides a chat but doesn't delete it.
• Deleted chats disappear from your view immediately and are scheduled for permanent deletion within 30 days, unless they were already de-identified, or OpenAI must keep them longer for security or legal reasons.
• Temporary Chats don't appear in your history, aren't used to train models, and don't create memories, but OpenAI may keep a copy for up to 30 days for safety purposes.
• Memory, if it's on, lets ChatGPT remember details from past chats and use them later. Turning memory off doesn't delete your past chats.

Is my data used for training?

On personal Free, Plus and Pro accounts, sharing for model training is on by default. You can switch it off under Settings, then Data controls, by turning off Improve the model for everyone (OpenAI's data controls page has the steps). You keep your chat history either way. Paying for Plus or Pro doesn't change that default. If you use ChatGPT without logging in, you can still turn the setting off, and the choice is saved in that browser. OpenAI says it doesn't train on business workspaces (Business, Enterprise, Edu) by default.

Who else can see my chats?

Mostly no one, with three exceptions worth knowing.

OpenAI's safety reviewers, in specific cases. OpenAI says that when it detects someone planning to harm others, the conversation goes to a small team of trained human reviewers who can ban accounts and, for imminent threats of serious physical harm to others, may refer it to law enforcement. It says it currently doesn't refer self-harm cases to law enforcement, out of respect for privacy.

Anyone you share a link with. Shared links can be forwarded. In 2025 OpenAI removed an opt-in option that let shared chats show up in search engines, after people found personal conversations in Google results, and it worked to get those removed.

Anyone with your device. Which brings us to the part most privacy guides skip.

The privacy risk nobody mentions: your partner

If you're using ChatGPT to process a fight, the person most likely to read it isn't a court. It's whoever picks up your unlocked phone, opens your laptop, or shares your account.

Think about:

• Your chat history. It's titled and searchable. "Is he gaslighting me?" sitting in the sidebar is a fight of its own.
• Synced devices. If you're signed in on a shared iPad or family computer, so is your history.
• Memory. If memory is on, ChatGPT might reference something you said about your partner in a later chat, possibly one they're looking over your shoulder at.
• Shared accounts. Couples sometimes share a subscription. Shared login means shared history.

What to do: use Temporary Chat for anything sensitive, delete chats you don't want to keep, turn memory off or review what's stored, and if your account offers locked chats, use them. And if you're afraid of what your partner would do if they read your chats, that's a sign to talk to the National Domestic Violence Hotline (1-800-799-7233 or text START to 88788), not a chatbot.

How to make your ChatGPT more private: settings to change today

Menus move around. If a setting isn't where this says, search OpenAI's help center for its name.

What not to paste into any AI

This applies to ChatGPT, Claude, Gemini, couples apps, and yes, us.

• Your partner's messages or screenshots. Those are their words. They didn't agree to send them to a tech company.
• Identifying details. Full names, addresses, workplaces, kids' schools.
• Health, legal or financial specifics you wouldn't want in a court file.
• Anything you'd hate to hear read aloud. That's the honest test.

You can still get good advice. "My partner said something dismissive about my job" works as well as pasting the text.

The APA's 2025 advisory on AI and mental health puts it bluntly: be extremely cautious about sharing sensitive information and avoid entering personally identifiable details.

5 questions to ask about any AI tool's privacy

Before you tell any AI about your relationship, check:

1. Is it saved, and for how long? Look for a specific retention period, not "as long as necessary."
2. Is it used for training, and can I turn that off?
3. Can I delete it, really? Look for permanent deletion, not just hiding.
4. Who else can see it? Human reviewers, other users, the other person in a shared tool.
5. What happens if someone sues? No consumer AI has legal privilege. A good company says so instead of implying otherwise.

How GoodTalk handles the "other person" problem

Two-person tools add a privacy question general chatbots don't have: will my partner see what I said? With GoodTalk, the AI mediator you text in iMessage, each partner's one-on-one thread stays private from the other. GoodTalk hears both sides but never shows one person's private messages to the other or quotes them without permission. If you want something passed along, you ask, and it's shared with clear attribution, and only what you asked to share. Like any AI tool, it isn't therapy and doesn't carry legal privilege.

For more on choosing tools, see Best AI for relationship advice, or our AI couples therapy guide for the bigger picture. And if you've been using ChatGPT like a counselor, Using ChatGPT as a therapist covers what it's good for.

What to do next

Open ChatGPT now and spend two minutes in Settings: turn off training if you'd rather not contribute, check what Memory has saved about your relationship, and delete anything you wouldn't want your partner to stumble on. Then make Temporary Chat your default for the hard stuff.