What should you never paste into an AI chatbot?
- Passwords, one-time codes and recovery phrases.
- Full card numbers and bank details.
- Passport, driving licence or national ID numbers.
- Other people's private information, such as a friend's medical details or a colleague's personal email.
- Confidential work material your employer has not approved for AI tools.
What happens to what you paste?
Your message is sent to the service's servers to be answered. Depending on the service and your settings, it may be stored in your history, kept for a period after you delete it, reviewed to improve the product, used to train future models, or passed to the AI provider that runs the model. None of that is sinister by default, but it means a chat is closer to an email than to a thought. The privacy policy and the data settings tell you which of these applies.
How do I share sensitive text more safely?
- Strip out names, addresses, account numbers and dates of birth. The AI rarely needs them to help.
- Use a temporary or incognito chat mode if the service has one.
- Check whether chats are used for training, and switch it off if you can.
- Delete chats you no longer need.
- Be wary of AI tools that act on websites for you. A page can hide instructions aimed at the AI, a risk called prompt injection, so keep agents away from accounts with a card attached.
A wider checklist is in is generative AI safe to use.
How does Ace handle what you share?
Your conversations with Ace are not used for targeted ads. Incognito Chat, switched on before your first message, is never saved, never in your history, and never used for memory or model training. Regular chats follow the privacy policy: data may be shared with cloud and AI providers and kept for up to 24 months after you stop using Ace. Ace also does not go to other sites and act for you, so it never types your details into a form on its own. More in does Ace train AI on my chats and on the safety page.