Does ChatGPT use my data for training?

On Free, Go, Plus and Pro, conversations may be used to improve models unless you turn off 'Improve the model for everyone' in Data Controls. The setting applies from then on, not to what was already used. ChatGPT Business, Enterprise and Edu, and the API, are not used for training by default.

Is ChatGPT confidential?

Not in the legal sense. Your chats are private from other users, but OpenAI stores them, may use consumer chats for training unless you opt out, and can be ordered by a court to preserve or produce them. In the New York Times lawsuit a federal judge ordered OpenAI to preserve user logs in May 2025, and later to produce 20 million de-identified logs. Do not put anything in a consumer account that you would not want to see in discovery.

Can a court get my ChatGPT conversations?

It has happened. On 13 May 2025 a federal court ordered OpenAI to preserve ChatGPT output logs, including deleted ones. The obligation was lifted going forward on 9 October 2025, but logs saved under the order stayed accessible and accounts flagged by the plaintiffs stayed under hold. In November 2025 the court ordered OpenAI to hand over 20 million de-identified logs, which was affirmed in January 2026. OpenAI said ChatGPT Enterprise, Edu and zero-data-retention API customers were not covered by the preservation order.

Is ChatGPT safe for confidential business information?

On a consumer account, no: assume it is stored, possibly used for training, and reachable by a court. On ChatGPT Business or Enterprise, or through the API, it is not used for training by default and retention is under your control, which is acceptable for most internal documents. For data you are not allowed to send to a third party at all, such as patient records or data under contract, the model has to run on infrastructure you control.

Does ChatGPT Store Your Data? Retention by Plan (2026)

Q: Does ChatGPT store your data?

Yes. On consumer plans (Free, Go, Plus, Pro) your chats stay in your account until you delete them. A deleted chat leaves your account at once and is scheduled for permanent deletion from OpenAI's systems within 30 days, unless it has been de-identified or OpenAI must keep it for legal or security reasons. Temporary Chats are kept up to 30 days for safety. Business and Enterprise workspaces follow their admins' retention settings.

A chat bubble sending documents along three paths: into a large open filing vault, into a safe that is closing on a timer, and into a small locked box inside a fenced company building

Short answer: yes, ChatGPT stores your data. How long, and whether it can be used to train models, depends on which ChatGPT you use. On the consumer plans (Free, Go, Plus, Pro), chats stay until you delete them, deletion takes up to 30 days, and conversations may be used for training unless someone turned that off. On ChatGPT Business, Enterprise and Edu, and through the API, nothing is used for training by default. And on any plan, a court can step in: in 2025 a federal judge ordered OpenAI to keep user logs it would otherwise have deleted. For a business, the real question is not "does ChatGPT store data" but which ChatGPT your team is pasting customer data into. Below is each case, and what we set up for clients who want AI without handing over their data.

What ChatGPT keeps, plan by plan

As described in OpenAI's help center articles on chat and file retention and data in consumer services, and on its enterprise privacy page. OpenAI changes these pages; check them before you rely on a detail.

Where your team types	Stored?	Used for training?	Deleting
Free, Go, Plus, Pro	Yes, until you delete	Yes by default, unless "Improve the model for everyone" is off	Gone from the account at once, purged within 30 days, unless de-identified or kept for legal or security reasons
Temporary Chat	Not in history; kept up to 30 days for safety	No	Automatic
ChatGPT Business	Yes	No by default	Purged within 30 days of deletion
ChatGPT Enterprise, Edu	Yes, for as long as the admin sets	No by default	Under the admin's retention settings
API	Kept for a limited period for abuse monitoring	No by default	Zero data retention available to eligible customers

Two details people miss:

The training opt-out is not retroactive. Turning it off stops future use. What was already used stays used.
"Delete" means "scheduled for deletion". The 30 days is how long OpenAI takes to purge, not a recovery window.

Courts can override "deleted"

Retention policies describe what a company does when nothing unusual happens. A lawsuit is unusual.

In the copyright case brought by the New York Times, a federal court in New York ordered OpenAI on 13 May 2025 to preserve ChatGPT output logs, including chats users had deleted. The order was lifted going forward on 9 October 2025, but logs saved under it remained accessible and accounts flagged by the Times stayed under hold (Yahoo News). In November 2025 the magistrate judge then ordered OpenAI to produce 20 million de-identified ChatGPT logs, and the district judge affirmed it in January 2026 (Data Privacy + Security Insider).

OpenAI said ChatGPT Enterprise, Edu and zero-data-retention API customers were outside the preservation order (OpenAI). Consumer and Team accounts were in it.

None of this means OpenAI is careless. It means that "private" in a chat app is a product setting, not a legal privilege. A conversation with a lawyer is privileged. A conversation with ChatGPT about the same contract is a business record held by a third party.

Ads, since 2026

ChatGPT began showing ads on the Free and Go plans in 2026. Reporting says advertisers do not see conversations, but ads are personalized with signals that include chat topics and memory, and OpenAI revised its privacy policy several times in the first half of the year as ads expanded (Search Engine Land). Paid plans, Business, Enterprise and Edu do not show ads. If your team uses free accounts for work, that is one more reason to stop.

What this means for a business

Most companies we talk to are not choosing between "AI" and "no AI". Their staff already use ChatGPT, mostly on personal accounts, because it saves them an hour a day. The data problem is that nobody decided where the customer list, the contract or the patient note is allowed to go.

Sort what your team handles into three buckets:

Public or harmless: marketing copy, public documentation, code that is already open source. Any plan is fine, with training off.
Internal and confidential: contracts, financials, customer emails, internal strategy. Use ChatGPT Business or Enterprise, or an internal tool built on the API, so nothing trains on it and your admin controls retention. Not personal accounts.
Data you may not send to a third party at all: health data, data covered by a client contract, anything a regulator says stays in-house. No hosted chat product fits, whatever its settings. The model has to run where the data already is.

Then make it easy to do the right thing. A policy that says "don't paste customer data into ChatGPT" without an approved alternative is a policy people will ignore by Thursday.

What we build for clients

We build the internal tool that replaces the personal ChatGPT account for a specific job, such as reading incoming documents, drafting replies, or updating the CRM after a call, so the data takes a path someone chose.

Through APIs, not consumer chat. Our AI features call models through APIs, under API terms: not used for training by default. We pick the model per task. That still means a provider processes the data, and we say so plainly to clients.

In-house when the data cannot leave. When it cannot, we run the model on infrastructure the client controls. Our own speaker diarization, for example, runs on GPU instances in our own Google Cloud project rather than through an outside AI API. The trade-offs, what it costs, and which option fits which case are on our sovereign AI page.

Access that ends when employment ends. People sign in with the Microsoft or Google account they already have, and rights come from their role. Remove someone and their access to the tool goes with it, which is not true of a personal ChatGPT account full of company documents.

Does ChatGPT Store Your Data? What Happens to What Your Team Pastes In (2026)

Short answer: yes. The useful answer depends on which ChatGPT your team is using, whether anyone turned off training, and whether a court has asked for the logs. Here is each case, and what we tell clients who want AI without handing over their data.