What happens to what you type into AI tools

When you type, paste, upload or speak something into an online AI tool, it leaves your device and goes to the provider's servers — often in another country. What happens next depends on the tool and, crucially, on which version you're using.

Free and personal accounts vs business accounts

	Personal / free versions	Business / enterprise versions
Used to train the AI?	Often yes by default, unless you opt out	Generally no, by contract
Read by people?	Possibly — some providers have human reviewers	Restricted under business terms
How long kept?	Can be months or years	Set by the company's agreement
Who controls it?	You, as an individual	Your employer's IT team

Some examples from the providers' own policies (as of 2025–26; they change often):

• ChatGPT: personal conversations may be used to train models unless you opt out; OpenAI says it doesn't train on business, enterprise or API data by default. (OpenAI)
• Google Gemini: a sample of personal chats may be read by human reviewers and kept for up to three years. Google itself advises not to enter confidential information you wouldn't want a reviewer to see. (Google)
• Claude: personal-plan users choose whether their chats can be used for training; business and API plans are excluded. (Anthropic)

The simple rule

Treat anything you put into a personal AI account as if you had posted it outside the company. Treat a company-approved AI tool like any other company system: safer, but still not a place for information you're not allowed to share.

What about tools that run on your device?

Some AI features — certain dictation, transcription and writing tools — process information on your device without sending it to a server. That's generally better for privacy, because the data never leaves your computer or phone. (Apple's explanation of on-device processing) But don't assume: check how a tool works, and still use only tools your company has approved.