Is AI safe? What happens to what you type

Where your chats go, what to keep out of them, and the two settings worth changing.

⏱ 6 min read · ▶ 1:12 video
The short version

Using a well-known AI assistant is about as safe as using any big website, as long as you treat it like a stranger who writes things down. What you type is sent to the company’s computers, may be stored, and on many free plans can be used to improve the AI unless you turn that off. So: no passwords, ID numbers, card numbers or private medical details.

📅 Checked as of October 2026. Apps and rules change, so check the sources below if something looks different.

Where your words actually go

When you type into ChatGPT, Gemini, Copilot or similar, your message travels over the internet to the company’s servers, where the AI writes the answer. Your chat is usually saved in your history so you can come back to it.

  • Training: companies may use conversations to train future versions. Most let you switch this off. In ChatGPT it’s a setting called Improve the model for everyone. In Gemini it’s tied to Gemini Apps Activity.
  • Human review: some companies say a small number of chats may be read by trained reviewers to check quality and safety. Google says not to enter anything you wouldn’t want a reviewer to see.
  • Temporary chats: ChatGPT’s Temporary Chat isn’t saved to your history or used for training (OpenAI may keep a copy for up to 30 days for safety checks).
Think of it like this: an AI chat is like talking to a helpful clerk at a big company’s front desk. Friendly and useful, but you wouldn’t read out your bank PIN, and you assume someone might keep notes.

Never type these into a chatbot

  • Passwords, PINs or the codes from sign-in texts
  • Social Security, passport or driver’s license numbers
  • Full card or bank account numbers
  • Someone else’s private details without their OK
  • Detailed medical records with your name attached (you can still ask general health questions)

Need help with a real document, like a confusing bill? Black out or delete the account number and your address first, then ask.

Other safety worries, in plain terms

  • It can be wrong. Chatbots sometimes make things up confidently. See why AI makes things up.
  • Fake AI apps. App stores have copycats. Get the official app from the company named in the listing (OpenAI, Google, Microsoft, Anthropic, Meta, xAI).
  • Scammers use AI too. Cloned voices and fake photos are real. See AI voice-clone scams and how to spot AI images and videos.
  • Kids and teens. Most chatbots require users to be 13 or older, often with a parent’s permission for under-18s. See AI, kids and homework.

How common is this now?

Very. In a Pew Research Center survey taken in February 2026 (published June 2026), 49% of U.S. adults said they use AI chatbots, up from about a third in 2024. In Pew’s survey published in September 2025, half of Americans said they were more concerned than excited about AI in daily life, and about six in ten wanted more control over how it’s used in their lives. Being careful is normal, not old-fashioned.

Quick tips

  • Turn off model training in your AI app’s settings if you’d rather your chats weren’t used. Here’s how in ChatGPT.
  • Use a temporary or incognito chat for one-off questions.
  • Delete old chats you no longer need.
  • Practice spotting AI mistakes in AI or Not?

Sources

Test yourself: play AI or Not? or Safe or Risky?, quick free games.

🛠️ Try it step by step

More in our How-To guides.

📖 Words to know

Chatbot Training data Generative AI AI companion