1. Stop AI apps training on your chats
Claude
Settings → Privacy → switch off “Help improve Claude”.
Turn it off
ChatGPT
Settings → Data controls → switch off “Improve the model for everyone”.
Turn it off
Gemini
Switch off “Keep Activity” in your Gemini activity settings. This also stops saving your chat history.
Turn it off
Instagram and Facebook
In the app: Settings → Privacy Centre → find the notice about Meta AI and tap “right to object”. Fill in the form to stop your posts being used for Meta’s AI. Meta has to confirm it by email.
Work plans (Claude Team and Enterprise, ChatGPT Business and Enterprise, Google Workspace) don’t train on your data by default. One more reason to use your work account for work.
2. Make your notetaker announce itself
Turn on its participant notice or chat message, so asking for consent happens by default, not by memory. Turn off auto-share to everyone on the invite.
The EU AI Act barely touches this, and GDPR’s answer is mostly obligations and fines for organisations. National law draws a hard line: For example, in Portugal, recording someone’s words without consent is a crime (Código Penal, art. 199).
3. Use a VPN on networks you don’t control
Public wifi, hotels, conference venues. Choose a paid VPN with an independent no-logs audit. A free VPN is often paid for with your data.
4. Lock the doors
A password manager for every login, and two-factor authentication wherever it’s offered.
5. Look after other people’s data too
Before you paste a client’s email, a colleague’s message or a contact list into an AI, ask: would they say yes?