AI Training On Your Data: Three Settings To Turn Off Today
Free and personal AI accounts use your chats for training unless you opt out. Here is how to switch it off in ChatGPT, Claude and Gemini, and which of the three deletes your history.
On free and personal accounts, AI training on your data is switched on by default and you have to turn it off yourself. On paid business accounts it is off by default. Either way the text leaves your computer and is stored on systems you do not control for some period of time.
August 2025, Claude. Free, Pro and Max moved from opt-in to opt-out. Training is on unless you turned it off, and conversations are kept five years if you left it on, thirty days if you turned it off.
The trend. All three big consumer AI services now train by default, so switching brands no longer avoids it. The setting, or a business account, is what protects you.
Checked September 2026. We re-check this every quarter.
Three Questions People Blur Together
Almost every argument about AI training on your data is really three separate questions with three different answers.
- Is it stored? Yes, on every tier, for some period.
- Is it used to train the model? This is the one the setting controls.
- Can a person read it? Sometimes, for safety review or quality checks.
When someone promises AI data privacy, ask which of the three they mean. They rarely mean all three, and only the middle one is what people picture when they worry about AI training on your data.
Consumer Tiers Versus Business Tiers
Whether AI training on your data happens at all comes down to which tier you are on. The defaults are reversed between consumer and business, and that reversal is the whole story.
| Tool | Personal or free | Paid business |
|---|---|---|
| ChatGPT | Free and Plus train by default. Opt out in settings. | Team, Enterprise and API do not train by default. |
| Claude | Free, Pro and Max train by default since August 2025. | Team and Enterprise do not train by default. |
| Gemini | Personal accounts train by default, and human reviewers may read some chats. | Paid Google Workspace business accounts keep Gemini chats out of training. |
Tell which one you are on in five seconds. Check which email address the account menu shows you signed in with. Work address usually means the company account. Personal address means a personal account, whatever computer you are sitting at.
3 Settings To Turn Off Today
If you use any of these on a personal account, this is worth two minutes each. Every path below was checked against the vendor's own help documentation.
On the web
- Open the account menu and select Settings
- Select Data controls
- Select Improve the model for everyone, turn it off, select Done
On mobile: open the sidebar, tap your profile icon for Settings, tap Data controls, turn off Improve the model for everyone.
It applies across every device. Your chat history stays. Source: OpenAI on how your data is used to improve model performance.
- Go to Settings, then Privacy
- Set Help improve Claude to off
This also shortens retention from five years to thirty days. If you have not touched this since August 2025, it is currently on.
- In Gemini, open Settings & help, then Activity (now called Keep Activity)
- Turn it off
Gemini is the one that costs you something. Turning that setting off also removes your chat history. ChatGPT and Claude let you keep your history and still stop training. Google ties the two together, so decide knowing that.
None Of This Is Retroactive
Every setting above stops AI training on your data going forward. Not one of them pulls back what has already gone in.
So for anything sensitive you already shared, flipping the switch now changes nothing. Tell us instead, the same day, using how to report a security incident. We would much rather know.
What A Business Account Does And Does Not Solve
A business tier stops AI training on your data. It does not make everything permissible.
The content still leaves your computer, still crosses the internet, still sits on someone else's systems under their retention schedule. Client agreements and professional obligations still apply on top. Those are separate questions from training, and they are answered by your company AI policy, not by the license.
Common Questions
Does turning this off make the tool worse for me?
No. It affects whether your conversations help train future models, not the quality of the answers you get now.
Is AI training on your data the same as a data breach?
No. It is disclosure you agreed to in the terms, which is a different problem with a different fix. The fix is the setting, or the right account tier.
What about the free tools built into other apps?
Assume consumer defaults until someone shows you the contract. An AI feature bundled into a free app is rarely covered by a business agreement.
Should I just delete my chat history instead?
Deleting is housekeeping. It does not reverse training that already happened, and on Gemini it is bundled with the setting anyway. Turn the setting off first, then tidy up.
How do I know which tools are approved here?
Ask us, or read your policy. If a tool is on the approved list, someone has read its contract. That is what approved means. Also see what you can paste into ChatGPT.
Not Sure Which Account You Are On? Ask Us.
Send us the tool name and the email address you sign in with, and we will tell you whether AI training on your data is switched on and how to turn it off. No AI data privacy question is too small to ask.