Most of us have typed something into ChatGPT, Gemini, or Copilot without giving much thought to where that information goes afterwards.
It feels like a private conversation, but it’s really more like using any other online service. The things you type, the files you upload, and sometimes even your location can be stored and used by these companies to improve their products. The good news is that all of them give you ways to limit this, and changing the settings takes about two minutes once you know where to look.
The safest general rule is to only share what the tool actually needs to do the job. You don’t need to put your full name, address or any sensitive details into a prompt for it to work. A general description usually does just as well. But if you do want to go further and tighten up the privacy settings on each platform, here’s how.
You can (and should) stop ChatGPT using your conversations for training.
By default, OpenAI can use your chats, prompts and any files you’ve uploaded to improve its models. To turn that off, open the ChatGPT app, tap your profile picture, go to Data Controls and switch off the option that says Improve the model for everyone. While you’re there, you can also clear your chat history from that same screen, and manage or delete anything ChatGPT has stored in its memory by going to your profile settings and selecting Memory, then Manage memories.
If you want even less saved going forward, there’s a Temporary Chat feature. Start a new conversation and tap the speech icon to open one. Memory is disabled for temporary chats, and they won’t show up in your history at all, which is handy if you’re asking about something you’d rather not have sitting in an account somewhere.
Google Gemini saves quite a bit unless you tell it not to.
Gemini can save your prompts and conversations to your Google account, and Google has said that a subset of chats are reviewed by human reviewers to check whether responses are low quality or harmful. That’s worth knowing if you’ve been assuming it was just an automated process. To limit this, open Gemini, tap your profile picture and select Gemini Apps Activity. Turn that off if you don’t want your conversations being stored.
You can also set up auto-delete from that same page, which will automatically remove saved activity after three, 18 or 36 months depending on what you choose. And if there are specific conversations you want gone now, you can scroll through your activity history and delete individual ones, or use the drop-down to wipe everything from the past hour, past day, or all time.
Copilot spreads its privacy settings across two different places.
Microsoft’s Copilot is a bit more scattered when it comes to privacy controls, with settings split between the app itself and your Microsoft account. Open the Copilot app, tap your profile picture, then go to Account and then Privacy. From there, you can turn off Training on conversation activity and Training on voice conversations if you don’t want your interactions used to improve Microsoft’s AI.
There’s also a Delete memory option on that same screen, which removes saved facts about you, customised instructions and anything Copilot has picked up from previous conversations. Worth knowing: this can’t be undone, so only do it if you’re happy to start fresh. You can also view, download or permanently delete your full conversation history from the same Privacy screen.
Claude has a straightforward setting for this.
Anthropic says that conversations, feedback, and coding sessions can be used to improve Claude, but turning that off is simple. Open the Claude app, tap your profile picture, select Privacy and switch off Help improve our AI models. That stops any new chats being used for training going forward.
It’s also worth going into Settings and checking Connectors, which shows you any third-party apps or services Claude has access to. Disconnect anything you don’t recognise or no longer use. And under Permissions you can see what Claude can access on your device, things like your location, calendar or photos, and switch off anything you’re not comfortable with.
Perplexity has an incognito mode which is actually quite useful.
Perplexity collects questions, prompts and uploaded content and can use them to improve its models, but you can turn that off by opening Settings, finding the toggle for AI Data Retention and switching it off. That stops future chats being used for training purposes.
The incognito mode is worth knowing about too. Go to Settings and turn on Incognito Mode, and anything you search in that mode won’t be saved to your account. The activity automatically expires after 24 hours, which is reassuring if you’re asking about something sensitive. You can also clear your full search history from the Settings page under the Security heading, and review which apps and services Perplexity has access to under Connectors.
A few general things worth keeping in mind
Even with all these settings adjusted, it’s worth remembering that AI tools can get things wrong. They can be out of date, confidently incorrect, or just not quite right for things involving your health, money or legal situation. Always double check anything important rather than taking it at face value.
You really don’t need to share much personal information to get useful results from these tools. Most of the time a general description of your question works just as well as a detailed personal one. Avoid uploading anything sensitive, like bank details, passwords or confidential work documents, unless you’ve read the privacy policy, and you’re comfortable with how that platform handles your data. A couple of minutes adjusting your settings across these apps is time well spent.
