Story6 min read

You tell an AI something personal. What happens to that conversation?

Many people ask AI what they don’t ask anyone else. Here’s what happens to those chats, and what you can do about it.

This box needs JavaScript. It runs only in this page.

    Nothing you type here leaves your browser.

    The example is made up. This check uses the same kind of word lists as Kamelo, and runs only on this page: nothing is sent, saved or measured.

    Why people ask AI

    It’s 2 a.m. You can’t sleep, and there’s something you haven’t told anyone. An AI is right there. It doesn’t judge, it doesn’t get tired, and it answers straight away.

    You’re far from the only one. In a survey of 300 ChatGPT users in the US, nearly half said they talk to it about health, and more than a third about money. 82% saw their AI chats as sensitive or very sensitive (AIES 2025).

    In a US survey for Tebra, 29% said they’re more comfortable talking about their mental health with ChatGPT than with their doctor (eMarketer). And in a survey by Proton of chatbot users in the US, the UK, France and Germany, 44% said they would share more if they were sure their chats would never be used to train AI (TechRadar).

    So people do share, and they’d like to feel safer about it. That starts with knowing what happens next.

    What happens to a normal chat

    • It’s in your history. The chat is saved under your account until you delete it. Its title sits in the sidebar, where anyone who sees your screen can read it.
    • The AI can remember it. ChatGPT, Gemini and Claude can remember things about you from one chat to the next. When ChatGPT says “Memory updated”, something you said was saved for later.
    • It can be used for training. Unless you turn it off, ChatGPT, Gemini and Claude can use your chats to improve their models.
    • It can shape the ads you see. The free versions of ChatGPT now show ads in many countries. An ad is picked based on what you’re talking about. Outside Europe, with ad personalization on, past chats and memory can count too. OpenAI says advertisers don’t get your chats, and that it doesn’t show ads next to health topics (OpenAI, Help Net Security).

    Two things have really happened. In the summer of 2025, thousands of shared ChatGPT chats could be found through Google, some with names and details about mental health and trauma. OpenAI removed the option that made this possible. Shared Claude chats have turned up in search results too (eWeek, Khaleej Times).

    And “deleted” wasn’t always deleted. Because of a lawsuit, OpenAI had to keep deleted and temporary chats from May to late September 2025. That order has ended, but what was kept in that period is still held, except for users in Europe and the UK (Simon Willison).

    Kamelo notices when you type something personal on an AI site, and offers a temporary chat instead. Free, no account. Add to Chrome

    What you can do yourself

    This takes about a minute per service, and it helps whether or not you ever use Kamelo.

    ChatGPT

    • Training off: Settings → Data controls → Improve the model for everyone.
    • Memory: Settings → Personalization. Delete what you don’t want it to keep under Manage memories, or turn off Reference saved memories.
    • Temporary chat: the Temporary button at the top right of a new chat.
    • Shared links: Settings → Data controls → Shared links → Manage.
    • Ads: Settings → Ad controls, to turn off ad personalization.

    Gemini

    • Training and history: Settings & help → Activity, and turn off Keep Activity.
    • Memory: Settings & help → Personal context, where your account has it. Turn off past chats, or delete what’s saved.
    • Temporary chat: the temporary chat button next to New chat.
    • Shared links: Settings & help → Your public links.

    Claude

    • Training off: Settings → Privacy → Help improve Claude.
    • Memory: Settings → Memory.
    • Incognito chat: the ghost icon at the top of a new chat.
    • Shared links: Settings → Privacy → Shared chats.

    Menus change now and then. If a name is slightly different, look for it in the same place.

    What Kamelo does, and doesn’t

    Kamelo is free privacy software for your browser. It works on ChatGPT, Claude, Gemini and other AI chats.

    • It notices. When you type something personal, like your health or money worries, Kamelo notices it on your own computer. It doesn’t send your text anywhere.
    • It offers a temporary chat. One click moves your question to a temporary chat in the same AI. That stays out of your history, isn’t added to memory and isn’t used for training.
    • In Chrome, it offers Melo. Kamelo’s private chat, an AI that runs in Chrome itself, on your computer. That question isn’t sent to any AI company. How Melo works.

    And what it doesn’t do:

    • It doesn’t make you anonymous. A temporary chat still sits under your account. The AI company knows it was you.
    • It doesn’t make chats vanish. ChatGPT and Claude keep temporary chats for up to 30 days for safety checks, Gemini for 72 hours.
    • It doesn’t hide names or details. The free version doesn’t change what you send. Masking is part of Kamelo Business.
    • Melo is Chrome only. In Edge, Kamelo suggests a temporary chat.

    Kamelo doesn’t want anything from you either: no account, and it’s open source, so anyone can check what it does.

    If you’re not okay

    An AI can help you put your thoughts in order. But it isn’t a doctor, a therapist or a crisis line, and it can be wrong. If you’re struggling, please talk to a person.

    Free, confidential support in your country: findahelpline.com. In the US, call or text 988. In the UK and Ireland, call Samaritans on 116 123. In an emergency, call your local emergency number.