In short
On personal ChatGPT, Claude and Grok accounts, your chats can be used to train their models unless you turn it off. We explain how to switch it off on each one, what the switch doesn't fix and why it pays to keep your data in a place of your own.
In this article
In short
By default, yes. On personal ChatGPT, Claude and Grok accounts, your conversations can be used to train their models unless you turn it off. It takes a minute to switch off in the settings, but it only applies to what comes afterwards. On business plans, all three say they don't train on your content by default.
What does each assistant do with your chats?
They work in a similar way: on personal accounts training is on and you decide whether to turn it off, as Engadget summarises.
- ChatGPT. OpenAI explains that on Free, Plus and Pro it may use your content for training. On Business, Enterprise and Edu it doesn't use instructions, answers or files, unless the organisation signs a separate agreement.
- Claude. Anthropic uses chats from Free, Pro and Max if you allow it, and in that case it may keep them de-identified for up to 5 years. On Team, Enterprise and the API it doesn't train by default.
- Grok. Its FAQ says it may use your messages, searches and what you upload, along with its answers, for training. With business customers it doesn't.
How do you turn it off?
- ChatGPT: Settings > Data controls > turn off “Improve the model for everyone” (OpenAI help).
- Claude: Settings > Privacy > turn off “Help improve our AI models” (Anthropic privacy centre).
- Grok: on grok.com, Settings > Data > “Improve the Model”; in the app, Settings > Data controls. If you use it inside X, the setting is in X's privacy settings.
For a sensitive conversation, all three have a mode that doesn't train: Temporary Chat in ChatGPT, incognito chat in Claude and Private Chat in Grok.
What doesn't the switch fix?
More than it seems:
- It isn't retroactive. Anthropic says it clearly: your data stays in training that is already under way and in models already trained. OpenAI and Grok only speak about new conversations.
- The thumbs-up counts. If you rate an answer with a thumbs-up or thumbs-down, OpenAI may use that whole conversation even if you have turned off training. Anthropic keeps the rated conversation for up to 5 years and may also use it; Grok does the same.
- It doesn't delete anything. Your chats stay in your history. Turning off training is not the same as deleting them.
What if what you give the AI is your real data?
One thing is what you write in the chat and another is the data the assistant queries. Anthropic clarifies that what it may use for training does not include the raw content of connectors, including MCP servers, unless you copy it directly into the conversation.
In other words: if your information lives in a place of your own and the assistant reads it through MCP when it needs it, you don't have to paste your spreadsheets into every chat. That said, what the assistant replies is part of the conversation, so the switch still matters.
How do we do it at Traiker?
At Traiker we start exactly there. Each person or company has their own encrypted database, and ChatGPT, Claude or Grok query it through MCP; we're opening access gradually in the beta.
- Your data is yours: we don't train models on it or process it.
- You export or delete everything whenever you want.
- You change assistants without moving your data, because it doesn't live inside any of them. We explain this in who owns your AI agent's memory.
If you're deciding what to share, review how to give your data to an AI without losing control.
Frequently asked questions
Does turning off training delete my history?
No. Your chats stay there; to delete them you have to remove them separately.
Do I lose the assistant's memory if I turn it off?
No. Memory and personalisation are in separate settings, so the assistant keeps answering you the same way.
Which account should my team use for company data?
A business one. On ChatGPT's Business and Enterprise plans, Claude's Team and Enterprise plans and Grok's business plans, their policies say your content is not used for training by default.
If you want your assistants to work with data that is still yours, sign up for the Traiker beta.
