Does ChatGPT Use My Data for Training? Privacy Settings Guide

Does ChatGPT use my data for training? On personal accounts, OpenAI says it may use your conversations to train its models unless you switch off the Improve the model for everyone setting, while its business plans and API are not used for training by default. As of 4 October 2026, Gemini, Copilot, Grok, DeepSeek and the free mode of Mistral also use consumer chats unless you opt out, and Claude says it trains on your chats only if you allow it. The table below summarises each vendor's setting, temporary chat mode and retention, taken from the vendors' own help and privacy pages.

Which AI chatbots use your chats for training?

Service and page date	Used for training by default?	How to opt out	Temporary chat	Retention
ChatGPT OpenAI help pages, updated in the week to 4 Oct 2026	OpenAI says it may use content from personal plans	Settings, Data controls, Improve the model for everyone, or the Privacy Portal	Temporary Chat: not used for training, not in history, deleted after 30 days	Kept until you delete; deleted chats removed within 30 days
Gemini Privacy Hub updated 15 July 2026	Yes while Keep Activity is on	Turn off Keep Activity in Gemini Apps Activity	Temporary chats: not used for training, kept 72 hours	Auto-delete defaults to 18 months; reviewed chats kept up to 3 years
Claude Anthropic pages dated 16 Mar and 1 Jul 2026	Only if you allow it in Privacy Settings, or if a chat is flagged for safety review	Switch off the model improvement setting	Incognito chats: never used for training	Deleted chats removed within 30 days; up to 5 years if you allow training
Copilot Microsoft pages, new app from 18 Aug 2026	Opt-out model for signed-in personal accounts	Privacy, Training on conversation activity and Training on voice conversations	Not described on the privacy pages we read	Conversation history kept 18 months by default
Grok SpaceXAI FAQ dated 12 May 2025	May use your content unless you opt out	Improve the model in Data controls (app) or Data (web)	Private Chat: not used for training, deleted within 30 days	Kept until you delete; deletion takes up to 30 days
Mistral (Vibe) Docs, undated	Free mode conversations can be used; Pro, Team and Enterprise are not by default	Admin Panel, Vibe, Privacy	Not described	Admins can set deletion from 30 days to 1 year
DeepSeek Terms, undated	Terms allow de-identified inputs and outputs to improve services	Turn off Improve the model for everyone	Not described	Personal data stored in the People's Republic of China

Pages for the same vendor can disagree on small points and change without notice, so treat the table as a map and confirm the setting in your own account.

How do you stop ChatGPT using your data?

OpenAI's data usage page says that on its services for individuals, such as ChatGPT and Codex, it may use your content to train its models. You have two ways out, and either one is enough.

1. On the web, click your profile icon, open Settings, then Data Controls, and turn off Improve the model for everyone. On mobile, open the sidebar, tap your profile icon and choose Data Controls.
2. Or select Do not train on my content in the OpenAI Privacy Portal.

The Data Controls FAQ adds details worth knowing. Opted-out chats still show in your history. A Temporary Chat is not used for training, does not appear in history or create memories, may be reviewed only to monitor abuse and is deleted after 30 days. If you save a temporary chat it becomes a regular chat and follows your settings. Giving a thumbs up or down sends the whole conversation to OpenAI, and it may be used for training even after you opt out. Codex has a separate setting for training on full environments.

What are Gemini's privacy settings?

Gemini's behaviour depends on the Keep Activity setting, described in Google's Gemini Apps Privacy Hub. With it on, chats are saved to your activity and used to provide and improve Google's services, including training generative AI models. A subset is read by human reviewers, including trained staff at service providers. Reviewed chats are disconnected from your account, kept for up to three years and not deleted when you delete your activity. Auto-delete defaults to 18 months, and you can change it to 3 months, 36 months or never.

Turn Keep Activity off and your future chats are not used to train Google's models unless you send feedback. They are still stored for 72 hours so Gemini can respond and Google can protect its services, and Google says human reviewers may still be involved in that safety work. Temporary chats are not used to train Google's models either and follow the same 72-hour rule. Google also says Gemini chats are not used to show you ads. Work and school accounts follow separate terms.

Developers should note a split in the Gemini API terms. Free usage, including Google AI Studio without billing, is used to improve Google's products and human reviewers may read it. Paid usage is not used to improve products. In the European Economic Area, Switzerland and the United Kingdom, the paid terms apply to all services.

How does Claude handle data retention?

For Free, Pro and Max accounts, Anthropic's training page (16 March 2026) says chats and coding sessions are used to improve Claude if you allow it, if a conversation is flagged for safety review or if you explicitly opt in, for example by joining a tester programme. The retention page (1 July 2026) describes a clear split. If you allow improvement, Anthropic may keep them in a de-identified form for up to five years, and this applies to new or resumed chats. If you turn the setting off, your previous and new chats are not used for future training. Data already inside a training run in progress or a trained model stays there.

• Deleting a chat: it leaves your history at once and is removed from back-end systems within 30 days.
• Incognito chats: not used to improve Claude, even with the model improvement setting on.
• Feedback: a thumbs up or down stores the related conversation for up to five years.
• Flagged chats: inputs and outputs that automated systems flag as policy violations are kept for up to two years, and classification scores for up to seven.

Anthropic presents the training setting as your choice, so open Privacy Settings and confirm it matches what you want rather than assuming.

What about Copilot, Grok, Mistral and DeepSeek?

• Copilot. The privacy controls page lets signed-in users opt out of training on conversations and voice conversations. The opt-out does not stop use for product improvement, advertising, safety, security and compliance. Work and school accounts are covered separately by Enterprise Data Protection.
• Grok. The SpaceXAI FAQ says feedback you give may be used for training even after you opt out, and that in some regions, outside the EU and UK, a signed-out user may have no opt-out. It asks users not to share personal or sensitive information.
• Mistral. The privacy docs say API data is not used for training, and that Labs models are an exception because data can be used regardless of plan or opt-out settings.
• DeepSeek. The privacy policy lists a right to opt out of training and says personal data is stored in the People's Republic of China. The terms of use name the toggle.

Are business and API plans different?

Yes, and the difference is the main reason companies pay for them. OpenAI says it does not use inputs or outputs from ChatGPT Business, Enterprise, Edu or its API to improve models by default. Anthropic says the same for its commercial products, including Claude for Work and the API, unless you report feedback or choose to share data. Mistral's Pro, Team and Enterprise plans are not used for training by default, and Microsoft routes work accounts through Enterprise Data Protection.

Not training is different from not storing. OpenAI's API data controls page says abuse monitoring logs, which can contain prompts and responses, are kept for up to 30 days by default. Zero Data Retention and Modified Abuse Monitoring need OpenAI's approval. Our guide to zero data retention for AI routers covers the API and business side in depth.

What should you never paste into a chatbot?

Treat every consumer chatbot as a place where a human reviewer could read your text. Google says so for Gemini, and xAI asks users not to share personal or sensitive information with Grok. Keep these out unless your employer has approved a business plan for them:

• Passwords, recovery codes, API keys and source code that contains credentials.
• National ID numbers, bank details and scans of identity documents.
• Medical information about you or anyone else.
• Client files, contracts under NDA and unreleased financial figures.
• Personal data about other people, such as names with contact details.

Practical habits help: use a temporary or incognito chat for one-off sensitive questions, replace names with placeholders, delete chats you no longer need, and read our guide on how to use AI effectively for safer prompting. If you need strong privacy for routine work, a model that runs on your own computer keeps the text on the device, and our comparison of the best AI model for each task lists local options.

Where Danube Labs fits

If your company wants AI tools with data handling you can explain to customers, we help choose and configure them. See our AI systems service or book a free intro call.