ShieldNook

Using AI assistants safely: what happens to what you type

AI assistants like ChatGPT, Gemini, Claude and Copilot are brilliant for drafting emails, planning trips and decoding confusing letters. But everything you type goes somewhere. Here’s where, and the settings that keep your private life private.

What happens to what you type

  • It’s usually saved. Most assistants keep your chats in your account history so you can come back to them.
  • It may help train future AI. Depending on the service and your settings, your conversations can be used to improve the company’s AI models.
  • People may read some of it. Google, for example, says a subset of Gemini conversations is reviewed by human raters, including trained service providers, to improve its services and safety.
  • It can stick around. Even after you change your settings, providers may keep chats for a limited time to run the service, keep it safe and meet legal obligations.

None of this means you shouldn’t use AI. It means you should treat a chatbot like a helpful stranger: friendly and useful, but not someone you’d hand your passport to.

Never paste these into a chatbot

  • Passwords, PINs, one-time codes or recovery phrases, including crypto wallet seed phrases.
  • Full card numbers, bank account details or government ID numbers.
  • Medical records, legal papers, or anything confidential from work, unless your employer has approved the tool.
  • Other people’s private information, like a friend’s address or a screenshot of someone else’s messages.
  • Photos of IDs, boarding passes or documents with barcodes.

Need help with a real document? Remove names, account numbers and addresses first, or swap them for placeholders like [NAME] and [ACCOUNT]. You’ll get the same help with none of the risk.

The privacy settings worth changing

Each major assistant lets you control whether your chats are used to train its AI. Here’s where to look:

  • ChatGPT: profile icon → Settings → Data Controls → turn off Improve the model for everyone. For one-off sensitive questions, use Temporary Chat: those chats aren’t saved to your history, aren’t used for training, and are deleted after 30 days.
  • Google Gemini: turn off Keep Activity (also called Gemini Apps Activity). Future chats won’t be used to train Google’s AI, though Google keeps them for up to 72 hours to run the service and keep it safe.
  • Claude: Settings → Privacy → turn off the model-improvement setting so new chats aren’t used to train future models.
  • Microsoft Copilot: profile icon → Privacy (under Settings or Account, depending on the app) → turn off Training on conversation activity and Training on voice conversations.

Apps change their menus often. If a setting has moved, search the provider’s help center for “model training” or “data controls”.

Watch out for these AI-era traps

  • Fake AI apps. Scammers publish look-alike “AI assistant” apps and browser extensions. Download only from the official app store, and check that the developer is the real company.
  • Too many permissions. AI tools that connect to your email, calendar, files or browser can see a lot. Connect only what you need, and review connected apps every few months.
  • Hidden instructions. AI tools that read web pages, emails or documents for you can be manipulated by hidden text written to trick them. Security researchers call this prompt injection. Don’t let an AI send emails, make purchases or share files without checking first.
  • Confident mistakes. AI can sound certain and still be wrong. Double-check anything important (medical, legal, financial or security advice) with an official source.
  • “AI support agents” that contact you first. If a chatbot or caller claims to be your bank’s or a tech company’s AI assistant and you didn’t reach out, treat it like any other unexpected message.

Kids and AI chatbots

Many AI assistants have minimum age requirements, so check before kids use one. Teach them the same rules: no personal details, no photos of themselves, and come and tell you if a chat gets strange or upsetting.

Your 5-minute privacy tune-up

  • Open the AI assistant you use most and find its privacy or data controls.
  • Turn off model training if you’d rather your chats stay out of it.
  • Delete old chats that contain personal details.
  • Review the apps and services connected to your AI account.
  • Use a temporary or incognito chat for sensitive one-off questions, where available.

AI assistants are some of the most useful tools ever made. Used wisely, they make you safer, not more exposed. That’s the spirit of ShieldNook Superintelligence .si. Next up: the new red flags of AI-written phishing.

Sources

Settings were checked against each provider’s help pages in September 2026. This guide is general information, not legal or professional advice.