How to use AI safely with your personal information

Understand the "postcard rule"

Before typing anything, pause and consider your message. The core idea for using AI safely is simple: treat everything you type into a chat window like writing on a postcard. If you wouldn't want a stranger reading it at the post office, do not send it to an AI. Most AI systems use your conversations for data training – this is where the AI company uses past chats to teach their computer models how to write and think better. Think of it like letting a student read your old letters so they can learn to write their own, but on a massive, anonymised scale.

You'll know it worked when you instinctively rephrase potentially sensitive information before hitting send. This habit helps the AI tool to assist you effectively without needing your private data. While AI can draft an email, it cannot magically remove personal information you've already typed.

Rephrase it to: "Help me budget a weekly salary of $1,500 for a typical household in a major city."

Organise your settings and opt out of training

Almost all major AI platforms allow you to turn off data training. When you switch this off, the company still processes your requests to give you an answer, but they promise not to use your chat history to teach future versions of their system. This setting usually applies to future chats, not ones you've already had.

To do this, open your AI tool. Look for a gear icon, your profile picture, or a menu with three horizontal lines, usually located in a corner of the screen (often bottom-left or top-right). Tap or click on this to open a menu. From the menu, find and select a button labelled "Settings", "Privacy", or "Data Controls". Within this section, look for a toggle switch or checkbox that says something like "Improve the model for everyone", "Allow model training", or "Use chat history for training". Toggle this option off. What happens next is a confirmation that your setting has been saved, or the toggle will simply show it's now off.

If it looks different: App layouts vary. If you don't see exact labels, explore sections related to 'Privacy & Safety', 'Data Management', or 'Chat History'. The key is to find an option that controls how your data is used for "model improvement" or "training". You might need to scroll down or look through a few sub-menus.

You'll know it worked when the toggle switch next to the "model training" option clearly shows it is in the "off" position (it might turn grey, or have a circle on the left side).

Sanitise your text before copying and pasting

"Sanitising" means cleaning your text to remove any specific details like names, phone numbers, email addresses, or financial figures before you feed them to the AI. This is your strongest defence against data leaks and ensures the AI doesn't learn sensitive facts about you. This action happens in your mind or a separate document, before you ever type into the AI chat.

When you review your text and replace real details with generic placeholders, the AI will still understand the context perfectly and can help you with your task. What happens is your input becomes general, focusing on the problem rather than the specifics of your personal life.

If it looks different: This step is about your pre-AI preparation, so what you see will be your own document, notes app, or even just your thoughts. No AI screen will look different here, as you're doing the sanitisation yourself.

Sanitised version: "Hi [Recipient Name], can you please transfer the [Refund Amount] refund for the project at [Project Address] to my [Bank Name] account by [Date]?"

You'll know it worked when you read through your prepared text and can't find any specific detail that would directly identify you, another person, or a specific location or account number.

Use temporary chats for sensitive tasks

If you need to discuss a particularly sensitive topic — like brainstorming a tricky work situation, or sorting out a personal issue — look for a temporary chat option. Some tools call this "Temporary Chat", "Incognito Mode", or "Ephemeral Chat". It works like write-on, wipe-off ink: once you close the window, the conversation is deleted from your history and generally won't be used for model training, even if your main training setting is on.

To activate this, open your AI tool and look near the top of your chat window or near the "New Chat" button for an option to start a temporary or incognito chat. Tap or click this before you begin your conversation. What happens is a new chat session will open, often with a visual indicator (like a different background colour or an "Incognito" label) to show it's temporary. This temporary chat is usually designed to be forgotten once you close it.

If it looks different: Not all AI tools offer a specific "temporary chat" feature, especially on free plans. If you can't find it, don't worry – simply ensure your main data training setting (from Step 2) is switched off, and always sanitise your information (Step 3).

You'll know it worked when you see a clear visual cue that your chat is temporary (e.g., a message like "This chat will not be saved," or a different interface colour).