Is ChatGPT Private?

The honest, current answer: no, not by default. Here's what OpenAI stores, what it trains on, the court order that froze deletions for millions of users, and which tiers are actually private.

The short answer

No. On a personal ChatGPT account, which covers Free, Plus, and Pro, OpenAI stores your conversations and uses them to train its models by default, unless you go into Settings and turn that off. Your chat history sits on OpenAI's servers tied to your account, and deleting a chat doesn't make it vanish on a schedule you control.

The picture changes a lot depending on which ChatGPT you're paying for. Business products, ChatGPT Team and Enterprise and the developer API, don't train on your data by default. And there's a 2026 wrinkle most people haven't heard about: a court has ordered OpenAI to preserve chats it would normally have deleted, including the ones users thought they erased. More on that below, with dates.

What "private" actually means

"Is ChatGPT private?" gets a fuzzy answer because the word is doing three jobs at once. Pull them apart and the question gets easy to answer.

Is it stored? This is about retention. Does a copy of your conversation live on OpenAI's servers after you close the tab, and for how long? On personal accounts, yes, indefinitely, as your account history.

Is it trained on? A separate question. Does your conversation get fed into the dataset that shapes the next model? This is the one most people mean when they ask if ChatGPT is "private," and on personal accounts the default answer is yes.

Can a person or a court reach it? Whether a human at OpenAI, or a litigant with a subpoena, can pull up what you typed. As long as the data exists somewhere, the answer is "under the right conditions, yes." That's exactly why the court order below matters.

Each of these has different rules per tier. Here's how they break out.

Personal accounts: Free, Plus, and Pro

This is what most people are using, and it's the least private posture OpenAI offers.

• Trains on your chats by default. Per OpenAI's Data Controls FAQ, conversations on personal accounts are used to improve future models unless you opt out.
• The opt-out is buried. It's at Settings → Data Controls → "Improve the model for everyone," toggled off. Most users have never opened that screen.
• Thumbs up or down overrides your opt-out. Even with training turned off, rating a response authorizes OpenAI to use that specific conversation for training. A small detail with a big footprint.
• History is retained. Anything you don't delete stays in your account. Deleted chats are scheduled for removal within about 30 days under OpenAI's standard retention policy, though see the court order section for the current exception.

Plus and Pro don't buy you a better data deal. The $20 and $200 tiers buy higher limits and better models. The training default and the retention behavior are the same as Free. People assume paying flips a privacy switch. It doesn't.

Business: Team, Enterprise, and the API

The deal flips once you move to OpenAI's business products.

• No training by default. OpenAI states it does not train on inputs or outputs from ChatGPT Team, ChatGPT Enterprise, or the API. For business tiers this is the contractual default, not a toggle you have to find.
• Tighter or configurable retention. Enterprise lets your admins set retention windows. The standard API keeps a short operational log for abuse monitoring rather than a permanent chat history.
• Zero Data Retention for the API. Eligible API customers can request a Zero Data Retention arrangement, where prompts and completions aren't stored at all after the request completes. This is the strongest posture OpenAI offers, and it mattered a lot in the court order.

So "is ChatGPT private" has a different answer for a solo user on Plus than for a company on Enterprise. Same brand, very different data handling.

Temporary Chat, explained honestly

OpenAI's Temporary Chat is the feature people reach for when they want a private conversation. It does real things, and it has real limits. Worth being precise.

What it does: per the Temporary Chat FAQ, the conversation stays out of your visible history, it isn't used to train models, and it's scheduled for deletion within about 30 days.

What it doesn't do: it doesn't make you anonymous to OpenAI. The message still travels to OpenAI's servers, still gets screened, and is still subject to any legal hold that suspends normal deletion. Temporary Chat is private from your own account history. It is not private from OpenAI. People confuse this with real incognito, and it isn't the same thing.

The court order sitting on your deleted chats

Here's the part that changed the answer in 2026, and it's the detail almost no marketing page mentions.

In May 2025, in the copyright case The New York Times v. OpenAI, a federal magistrate judge ordered OpenAI to preserve ChatGPT output logs it would normally delete, including deleted chats and Temporary Chats. The plaintiffs argued those logs were potential evidence. OpenAI's routine 30-day deletion was paused for a huge swath of users while the litigation plays out.

Not everyone was swept in. ChatGPT Enterprise, Edu, and API customers with a Zero Data Retention agreement were carved out of the preservation order. The carve-out is the clearest signal you'll find of which tiers OpenAI itself treats as genuinely no-retention.

It escalated. On January 5, 2026, U.S. District Judge Sidney Stein affirmed a related order requiring OpenAI to hand over a sample of 20 million de-identified ChatGPT conversations to the plaintiffs, per Bloomberg Law. OpenAI has said the data is restricted to a small audited legal and security team and won't be handed to the Times directly, which you can read in its own response to the NYT data demands.

Privacy by tier, side by side

Find the row you're actually using. The right column is what's happening to your data right now.

So, is it private?

If you're on a personal account and you've never touched Data Controls, the honest answer is no. Your chats are stored, they're training the next model, and as of 2026 even your deleted ones may be frozen in place by a court. Turning off training and using Temporary Chat helps, but it doesn't change the underlying fact that OpenAI is in the data path and holds what passes through.

If you're on Enterprise or the API with Zero Data Retention, you're in much better shape, which is why those tiers were the ones a federal court agreed to leave alone.

The most private setup isn't a tier of ChatGPT at all. It's an architecture where no one keeps the conversation. You bring your own key, the messages go straight to the model, and nothing is stored on the way through. That's the design behind Private Claude, and it's the same reasoning that explains whether Anthropic reads your chats and how Claude Pro compares to Private Claude on exactly this question.