What should you never tell ChatGPT?
Never paste passwords, card or bank numbers, customers' personal data, health records, unreleased financials or anything under an NDA into a personal AI account. On personal plans, switch off model training in settings, and treat a work AI account like work email: your employer may be able to read it.
AI apps are safe enough for most business work. The risk is not the app. It is what you paste into it.
Never paste these into a personal account
- Passwords, API keys, login codes. Not even “just to format them.”
- Card numbers and bank details. Yours or anyone else’s.
- Customers’ personal data. Names plus emails, phone numbers, addresses, order histories. Strip them out first or use placeholders like [Customer A].
- Health or legal records about real people.
- Unreleased numbers. Revenue, payroll, a deal that is not signed.
- Anything under an NDA. If you signed to keep it private, an AI app counts as telling someone.
- Other people’s private messages without their OK.
Switch off training on personal plans
| App | Where | What to switch off |
|---|---|---|
| ChatGPT | Settings, Data Controls | “Improve the model for everyone” |
| Claude | Settings, Privacy | “Help improve Claude” |
Business, Team and Enterprise plans of both are not used for training by default. That is one of the real reasons to pay for a team plan once staff use AI.
Can your employer see your chats
On a company AI account, assume yes. On Claude Team and Enterprise, the organization’s Primary Owner can export all conversation data (Claude help center). Other enterprise AI plans often keep records for compliance. Ask for your company’s AI policy, and treat it like work email.
If you run the business: tell your team this plainly, in writing. It builds trust and stops someone pasting the customer list into their personal account.
The 10-second test
Before you paste, ask: would it hurt if this showed up somewhere it should not? If yes, strip the names and numbers, or do not paste it. Everything else, go ahead. That covers most of the real work.
People also ask
Can my company see what I do with Claude or ChatGPT?
On a company account, yes, it can be possible. On Claude Team and Enterprise, the Primary Owner can export the organization's data, which includes conversations. Enterprise plans from other vendors often keep chat records for compliance too. Use a personal account for anything you would not put in a work email.
Does ChatGPT train on my conversations?
On personal plans (Free, Go, Plus, Pro), yes by default. Turn it off in Settings, Data Controls, 'Improve the model for everyone.' ChatGPT Business, Enterprise and the API are not used for training by default.
Does Claude train on my conversations?
On personal plans (Free, Pro, Max), Anthropic asks you to choose, and you can switch it off any time under Settings, Privacy, 'Help improve Claude.' Team, Enterprise and API use are not used for training.
What should you never say to AI?
Anything you would not want read out in a meeting: passwords, account numbers, other people's private details, and secrets you signed an agreement to keep. The rule is simple: if it would hurt to leak, do not paste it.