This guide shows the one setting to change in each app, plus the habits that keep business data safe.
We get asked about this a lot. Somebody pastes a client email into a chatbot to tidy it up, and then wonders where that email went. The honest answer is that it depends on which plan you're on and which boxes you left ticked.
The short version
Do company work on a business plan or a paid API account. Those don't train on your data by default. On any personal account, turn training off before your first chat.
Three things can happen to what you type
People tend to lump these together, but each one has its own settings.
| What happens | What it means | Can you turn it off? |
|---|---|---|
| Training | Your chats help build future models. Bits of what you typed could resurface in ways nobody can predict. | Yes, on every major chatbot. It's off by default on business plans and APIs. |
| Retention | The company keeps your chats on its servers for a while, even if it never trains on them. | Partly. Deleting chats and using temporary or incognito chats shortens it. |
| Human review | Staff may read chats that get flagged for safety, or chats you send as feedback. | Mostly no. You can avoid sending feedback on anything sensitive. |
Turning training off doesn't make a chat private. It stops one use of it. Client data still belongs on a business account.
Where the training switch lives in each app
| AI tool | Setting to turn off | Where to find it | Business plan trains on your data? |
|---|---|---|---|
| ChatGPT | Improve the model for everyone | Settings, then Data Controls | No, for Business, Enterprise, Edu and the API |
| Claude | Help improve Claude | Settings, then Privacy | No, for Team, Enterprise and the API |
| Google Gemini | Keep Activity | Gemini Apps Activity at myactivity.google.com | Work and school accounts fall under separate Workspace terms. Paid Gemini API projects aren't used to improve Google products. |
| Grok | Improve the Model, plus a separate box in X | grok.com Settings, then Data Controls. In X, Privacy and safety, then Grok & Third-party Collaborators | No, for Grok Business, Enterprise and the xAI API |
| Meta Muse | Help improve our AI models | Settings, then Data Controls | Muse has no business plan yet |
Habits that matter more than any setting
Settings get reset. Plans change. A flagged chat can end up in front of a reviewer. So a few things should never go into any AI chat, whatever you've switched off.
- Never paste passwords, API keys, MFA codes, private keys or connection strings. If one slips in, change it.
- Keep client data, health information, card numbers and Social Security numbers out of personal accounts entirely.
- Swap real names, IP addresses and usernames for placeholders before asking for help with a log or a script.
- Sign in with your work account for work, so your IT team can manage the settings and remove access when someone leaves.
- Only connect the mailboxes, drives and apps a task needs, and disconnect them afterwards.
- Skip the thumbs up and thumbs down buttons on sensitive chats. Feedback is often kept for years and can be used for training even when training is off.
Quick checklist
Run through this once per account, then again every quarter.
- Company work happens on a business plan or paid API, not a personal account.
- The training setting is off on every personal account: ChatGPT, Claude, Gemini, Grok on grok.com and in X, and Muse.
- Temporary or incognito chats get used for anything sensitive.
- Old shared chat links are deleted.
- Nobody has opted into a "share data for free credits" program on an API account.
- No credentials or regulated data in any chat, ever.
Frequently asked questions
Does turning off training delete what I've already shared?
No. It only applies going forward. Anything already used for training stays in that model. Deleting old chats keeps them out of future training runs.
Are paid personal plans like ChatGPT Plus or Claude Pro private by default?
No. Paying for a personal plan doesn't change the training default. You still need to switch the setting off yourself.
Which AI chatbot is safest for business data?
Any of the big ones, as long as you use its business plan or a paid API account and configure it properly. The plan type matters far more than the brand.
Want help rolling these settings out across your team, or picking the right business AI plan? Talk to your ECS account manager and we'll sort it out with you.
Privacy guides for other AI tools
- How to stop ChatGPT from training on your data
- How to stop Claude from training on your data
- How to stop Google Gemini from training on your data
- How to stop Grok from training on your data
- How to stop Meta Muse from training on your data
- AI chatbot privacy settings: the overview and checklist
Last checked October 2026. AI companies rename and move these settings often, so if a menu looks different, search the app's settings for "training" or "improve".