AI Chatbot Privacy Settings: ChatGPT, Gemini, Claude
Where to find training, history, and deletion controls in ChatGPT, Gemini, Claude, and Microsoft Copilot, and what each setting changes about your chat data.
Short answer
Each major chatbot has a setting that controls whether your chats are used to train its models, plus options to pause or delete chat history. Open the privacy or data controls section of your account, turn off model training if you prefer, and use temporary or incognito chats for anything sensitive.
Key takeaways
- ChatGPT, Gemini, Claude, and Copilot all let signed-in users stop future chats from being used for model training.
- Turning training off does not always mean chats vanish at once; services keep some data for a short period for safety and abuse checks.
- Temporary or incognito chat modes are the simplest way to keep a single sensitive conversation out of your history.
Chatbots remember more than most people expect: the conversation itself, files you upload, and sometimes details they save for later replies. Every major service now gives you controls over that data, but the settings have different names and the menus move around between app updates. This guide explains what to look for in ChatGPT, Gemini, Claude, and Microsoft Copilot, and what each control actually changes, based on each company's own help pages as of November 2026.
Which privacy settings should you check first?
Check three things on every chatbot: whether chats are used for model training, how long chat history is kept, and whether a temporary or incognito mode is available.
These are separate controls, and turning one off does not change the others. A common mistake is assuming that disabling training also deletes history. It does not.
| Service | Training control | History and deletion | Private mode |
|---|---|---|---|
| ChatGPT | "Improve the model for everyone" in Data controls | Delete single chats or all chats; history kept separately from the training setting | Temporary Chat |
| Gemini | Keep Activity (Gemini Apps Activity) | Auto-delete after 3, 18 (default), or 36 months, or off | Temporary chats |
| Claude | Model Improvement in Privacy Settings | Delete individual conversations | Incognito chats |
| Copilot | Model training option in Copilot privacy settings | Delete single conversations or the full history; stored 18 months by default | Varies by app version |
Setting names come from each company's help pages. Labels can change with app updates, so if a name does not match exactly, look for the closest option in the privacy or data section.
How do you review ChatGPT data controls?
In ChatGPT, open Settings, go to Data controls, and switch off the model-improvement option if you do not want your chats used for training.
OpenAI's help center says data sharing for training is on by default for personal Free, Plus, and Pro accounts and that you can turn it off while keeping chat history. When you are signed in, the choice applies to your account on web and mobile.
For a single sensitive conversation, start a Temporary Chat. OpenAI says these chats stay out of your history, do not create or update memories, and are not used to train models, though they may be kept for up to 30 days for safety. If you save a temporary chat, it becomes a regular chat and follows your normal settings. ChatGPT's memory feature is a separate control; review saved memories in Settings and delete anything you did not mean to share.
How do you manage Gemini activity?
For Gemini, open the Gemini Apps Activity page from your Google Account or the Gemini menu, then turn Keep Activity off or set a shorter auto-delete period.
Google's Gemini Apps Privacy Hub explains the trade-off clearly. With Keep Activity on, chats and uploads are saved and may be used to improve Google services, including training generative AI models, with help from human reviewers. With it off, future chats are not saved to your activity or used for training (unless you send feedback), but Google keeps them for 72 hours to respond and protect the service.
Two details are easy to miss:
- Chats already reviewed by humans are disconnected from your account and kept for up to three years, even if you later delete your activity.
- Auto-delete is set to 18 months by default. You can change it to 3 or 36 months.
How do you control Claude's data use?
In Claude, open Settings, go to the Privacy section, and set Model Improvement to off if you do not want your chats used for training.
Anthropic's privacy center says that for Claude Free, Pro, and Max accounts, chats may be used to improve models if Model Improvement is on, if a conversation is flagged for safety review, or if you opt into programs such as trusted testing. Incognito chats are not used to improve Claude even when Model Improvement is enabled.
Feedback is handled differently. If you rate a response with a thumbs up or down, Anthropic says that conversation may be stored for up to five years, de-linked from your user ID. Skip the rating buttons on anything confidential.
How do you change Microsoft Copilot privacy options?
In Copilot, sign in, open your profile or settings menu, find the privacy section, and turn off model training; you can also delete individual conversations or your whole history there.
Microsoft's Copilot privacy FAQ says the training opt-out is available only to signed-in users, and opting out stops future conversations from being used for training while personalization can stay on. Conversation activity is stored for 18 months by default, and uploaded files are deleted automatically within 18 months.
Microsoft also notes a migration deadline: an updated Copilot app arrived in August 2026, and people using the older version need to sign in to the updated app by December 31, 2026, or unmigrated content will be deleted. If you rely on old Copilot chats, export or move them before then.
What else keeps your chats private?
Beyond the settings, the most reliable protection is not pasting secrets into a chatbot at all.
A short checklist:
- Remove names, account numbers, and passwords from documents before uploading.
- Use temporary or incognito modes for medical, legal, or financial questions.
- Review connected apps (email, cloud drives, calendars) and disconnect any you no longer use.
- Secure the account itself with a strong password and two-factor authentication, since anyone who gets in can read your history.
- Recheck settings after major app updates, because new features sometimes arrive with their own toggles.
If you want prompts that never leave your device, a local model is the alternative; how to run AI models locally explains the hardware you need. And because anything a chatbot tells you back may be wrong, keep how AI hallucinations happen in mind before acting on its answers.
Frequently asked questions
If I turn off model training, are my chats deleted?
No. The training setting only controls whether future chats can be used to improve models. Deleting chats is a separate action, and some services keep data for a limited time after deletion for safety purposes.
How long does Gemini keep my chats?
Google's Gemini Apps Privacy Hub says activity is auto-deleted after 18 months by default, and you can change that to 3 or 36 months or turn auto-delete off. Chats made with Keep Activity off are still kept for up to 72 hours.
Do work or school accounts follow the same rules?
Usually not. Business, enterprise, and education plans often have different defaults set by the organization's contract and admin. OpenAI, for example, says it does not train on ChatGPT Business, Enterprise, or Edu workspace data by default.
Can people at the AI company read my chats?
In limited cases, yes. Google says a subset of Gemini chats is reviewed by human reviewers when Keep Activity is on, and OpenAI says human review can happen in limited circumstances, including safety cases, even with model improvement turned off.
Sources
- Data controls in ChatGPT (OpenAI Help Center)
- Gemini Apps Privacy Hub (Gemini Apps Help)
- Is my data used for model training? (Claude Privacy Center)
- Privacy FAQ for Microsoft Copilot (Microsoft Support)
Research-based: Technorat does not claim hands-on testing. See how we research. Spotted an error? contact@technorat.com