AI Learning Hub
beginner

Privacy & where your data goes

When you talk to ChatGPT or Claude, where does your text actually go? What happens to it? And what should you never paste in?

What's actually happening

When you send a message to ChatGPT, Claude, or Gemini through their free chat product:

  1. Your message leaves your browser as encrypted data over the internet (HTTPS — the lock icon).
  2. It arrives at the company's servers, which decrypt it and feed it to the model.
  3. The model produces a reply.
  4. Your message and the reply are typically logged — saved to the company's database.
  5. Depending on the product, those logs may be used to:
    • Train future models (most free products by default; usually opt-out-able).
    • Be reviewed by humans for safety / abuse / quality reasons.
    • Be retained for some period (often 30 days to indefinitely).

For paid / enterprise / API customers, the contract is different — usually no training on your data, shorter retention, sometimes data residency commitments (your data stays in a specific country/region).

For on-device AI (Apple Intelligence on your phone, local models you run yourself), the data may never leave your device at all.

Real-world examples

A practical decision tree

Two questions decide how much privacy guardrail you need before pasting anything into a chatbot.

Try it yourself (no coding, ~3 minutes)

Find your data settings right now:

  1. ChatGPT: Settings → Data Controls. Toggle off "Improve the model for everyone" if you want.
  2. Claude.ai: Settings → Privacy. Check the current data retention setting.
  3. Gemini: Settings → Activity. You can pause activity tracking, set auto-delete.
  4. Microsoft Copilot: Privacy settings vary by product (free Copilot vs. Microsoft 365 Copilot are different).

Even if you don't change anything, knowing where the toggles are puts you ahead of 99% of users.

Check your understanding

  1. 1. You want to ask Claude.ai to debug a code snippet that includes an API key from your work. The right move is:
  2. 2. Which is the strongest privacy posture for AI?
  3. 3. What's the privacy advantage of running an AI model on your own laptop?

Found this useful? Share it with someone learning AI.

Further reading

Related lessons in this track