“Private” means two different things, and conflating them is why most answers to this question are wrong.
Your chats are private from other users — nobody else can browse your history. They are not private from OpenAI, which stores them, can review some of them, and until recently was under a court order to preserve them.
Here is what the policies actually say, with the retention periods spelled out.
📋 Key Takeaways
- Deleted chats disappear from your view at once and are purged from OpenAI systems within 30 days
- Training on your conversations is on by default for Free, Plus and Pro — it is a setting, not a plan feature
- Turning training off is not retroactive: anything already in a completed training run stays there
- Temporary Chat is the strictest consumer option — deleted within 30 days, never used for training
- A VPN hides your IP address. It does not stop OpenAI from storing what you type
What OpenAI Keeps, and for How Long
| What | Retention | Used for training? |
|---|---|---|
| Active conversations | Until you delete them | Yes, by default |
| Deleted conversations | Removed from view immediately; purged within 30 days | No |
| Temporary Chats | Deleted within 30 days | No |
| Archived chats | Kept until deleted — archiving is not deleting | Yes, by default |
| Chats flagged for policy review | Retained beyond 30 days | Reviewed by humans |
Source: OpenAI Help Center — Chat and file retention policies and Data Controls FAQ. Last checked 22 July 2026.
The distinction people miss most often is archive vs delete. Archiving only removes a chat from your sidebar. It stays in your account, and it stays in scope for training.
The Default Setting Nobody Changes
On Free, Plus and Pro, ChatGPT uses your conversations to improve OpenAI’s models unless you say otherwise.
To turn it off: Settings → Data Controls → “Improve the model for everyone” → off.
Two caveats that matter more than the setting itself.
It is not retroactive. Data already used in a completed training run cannot be pulled back out. The toggle governs future conversations only.
It does not stop storage. Your chats are still saved to your account and still retained. Opting out of training is not opting out of retention — those are separate things, and vendor help pages rarely make the difference obvious.
What the New York Times Lawsuit Changed
In May 2025, a magistrate judge in the Southern District of New York ordered OpenAI to preserve output log data that would otherwise have been deleted — including chats users had already deleted.
For roughly a year, “deleted within 30 days” was not true for many users.
That broad mandate has since been lifted. OpenAI has returned to standard retention, and the order no longer requires indefinite preservation of new conversations, or of conversations originating in the EEA, Switzerland or the UK. The Times continues to seek a specific set of data from April–September 2025.
Source: OpenAI — Response to NYT data demands. Last checked 22 July 2026.
The durable lesson is not about one lawsuit. It is that a retention policy is a statement of intent that litigation can override. Anything you typed into ChatGPT in 2025 may still exist because a court said so, regardless of what the help page promised at the time.
Business and API Tiers Are Different
Consumer-tier answers get misapplied to work accounts constantly. The terms are not the same.
| Tier | Trained on by default | Notes |
|---|---|---|
| Free / Plus / Pro | Yes | Controlled by Data Controls toggle |
| Team / Enterprise / Edu | No | Business data excluded by default |
| API | No | Not used for training without opt-in |
If you are pasting client material into a personal Plus account, you are on the top row — not the bottom two.
Where a VPN Helps, and Where It Does Not
This site earns affiliate revenue from VPN services, so let me be precise rather than flattering.
A VPN does not make your ChatGPT conversations private. You are logged into an account. The content travels to OpenAI, gets stored against your user ID, and is subject to the retention table above. Routing the connection through another country changes none of that.
What a VPN actually does:
- Hides your IP address and approximate location from the service
- Stops your ISP or network operator from seeing which sites you connect to
- Encrypts traffic on networks you do not control, like hotel or airport Wi-Fi
- Lets you reach services that are unavailable on your current network
That last one is the real use case for most readers here — see our guide to reaching AI tools from restricted networks. If your goal is account-level privacy from the AI vendor, a VPN is the wrong tool; the right tools are Temporary Chat, the training toggle, and not typing the sensitive thing in the first place.
If you do want one for the network-level reasons above, ZoogVPN and Proton VPN both publish no-logs policies and offer free tiers you can test before paying.
Practical Settings, in Order of Impact
- Use Temporary Chat for anything sensitive. Strictest consumer option: no training, 30-day deletion, no history.
- Turn off “Improve the model for everyone” in Data Controls. Future chats only.
- Delete rather than archive. Archiving keeps the chat live in your account.
- Turn off Memory if you do not want details carried between conversations.
- Use a work-tier account for work data. Team and Enterprise exclude your content from training by default.
- Redact before you paste. Names, account numbers and credentials do not need to be in the prompt for the model to help you.
How ChatGPT Compares to Other Assistants
Retention policies vary more than most people expect. Google keeps human-reviewed Gemini conversations for up to three years even after you delete your activity. Anthropic keeps conversations for five years if you leave training enabled, and 30 days if you opt out.
We put every vendor’s stated numbers side by side in which AI chatbot is actually the most private.
Also see: Claude vs ChatGPT 2026 · ChatGPT alternatives · AI Tool Finder
FAQ
Is your data safe with ChatGPT?
Safe from other users, yes — nobody can browse your history. Safe from OpenAI, no. Conversations are stored against your account, may be reviewed by humans if flagged, and are used to improve models by default on consumer plans.
Can someone see my ChatGPT history?
Not other users. OpenAI staff can access conversations in limited circumstances, including abuse investigations and legal requirements. Anyone with access to your account can see your history, so account security matters more than most privacy settings.
Does ChatGPT store my data forever?
No. Deleted conversations are removed from OpenAI’s systems within 30 days under standard policy. The exception is legal preservation — during the New York Times litigation, OpenAI was ordered to retain logs it would otherwise have deleted.
Should I be careful what I tell ChatGPT?
Yes. Avoid credentials, financial account numbers, client-confidential material, and health details you would not put in an email to a vendor. Use Temporary Chat when the topic is sensitive.
Does deleting my ChatGPT account delete my data?
Account deletion removes your conversations and is not reversible. It does not remove data already incorporated into completed model training runs, which cannot be reversed by deletion.
Does a VPN make ChatGPT private?
No. A VPN hides your IP address from OpenAI and hides your traffic from your ISP. Your conversations are still tied to your logged-in account and stored under the same retention policy.