The short answer: never share passwords and codes, ID and bank details, other people’s private information, confidential work information or sensitive health details with an AI assistant. General questions, public information and your own writing with names removed are usually fine. Before you send, ask: “Would I be okay if this ended up on a postcard?”
AI assistants like ChatGPT, Gemini, Claude and Microsoft Copilot can help with almost anything you type in. That’s the point, and also the risk. What you share may be stored, and depending on your settings, it may be used to improve the service. A good rule: treat what you type into AI like writing on a postcard.
Why it matters: what happens to what you type
When you chat with an AI assistant, your messages are sent to the company’s servers. What happens next depends on the tool and your settings, but the companies are open about the basics:
- Your chats may be used to train the AI. For personal ChatGPT accounts, OpenAI says sharing your data to improve its models is on by default, and you can turn off the setting called “Improve the model for everyone” (OpenAI Help Center).
- People may read some chats. Google’s Gemini privacy page says human reviewers read some of the data it collects, and asks: “Please don’t enter confidential information that you wouldn’t want a reviewer to see or Google to use to improve our services” (Gemini Apps Privacy Hub).
- Mistakes can be costly. In 2023, Samsung banned staff from using ChatGPT and similar tools after an engineer uploaded sensitive internal source code (Forbes).
None of this means AI is unsafe to use. It means you should decide what goes in, the same way you would with an email to someone outside your company.
Never share these
| Type of information | Examples |
|---|---|
| Passwords and codes | Login details, one-time codes, PINs |
| ID and money details | ID or passport numbers, bank and card numbers |
| Other people’s private details | A colleague’s phone number, a customer’s address |
| Confidential work information | Contracts, unreleased plans, client data |
| Sensitive health details | Medical records with your name on them |
A good test for each one: if a stranger read it, could it hurt you, someone else or your organization? If yes, keep it out.
What’s usually fine
- General questions: “How do I write a polite reminder email?”
- Your own writing with names and details removed.
- Public information, like a news article or a product description.
- Topics you want to learn about, like “Explain how a mortgage works.”
- Your own ideas and plans, described in general terms.
Build safe AI habits
Step 5 of AI Essentials covers exactly what never to share, and why.
Get great help without the private bits
You rarely need real details to get a useful answer. Swap them for placeholders:
Rewrite this email to a client so it sounds more confident: “Hi [Client], the project will be ready on [date]…”
The AI doesn’t need the real name or date to fix the tone. You fill them back in afterward.
Before and after
Too much:
My colleague Sarah Tan (phone 9123 4567) keeps missing deadlines on the Acme Bank contract worth $250,000. Help me write an email to her.
Just enough:
A colleague keeps missing deadlines on an important client project. Help me write a kind but clear email asking how I can help and agreeing on new dates.
The second prompt gets you the same useful email, without sharing a name, a phone number, a client or a contract value.
How to clean text before you paste it
- Copy it into a blank document first, not straight into the AI.
- Replace names with roles: “Manager,” “Customer,” “Colleague A.”
- Delete numbers that identify anyone: phone, ID, account, order and policy numbers.
- Remove signatures, addresses and email footers.
- Cut anything you don’t need help with. Less text means less risk.
For long emails and reports, see our guide to summarizing with AI safely.
Check your settings
Most AI assistants let you control your chat history and whether your chats are used to train the service. Take two minutes to look in the settings of the one you use. In ChatGPT, for example, look under Settings, then Data controls. At work, follow your organization’s rules on which AI tools are allowed. Many organizations provide work versions of AI tools with stronger privacy protections.
Using AI at work: 4 questions to ask first
- Is this tool allowed? Many organizations have a list of approved AI tools. Using a personal account for work can break the rules even if your intentions are good.
- Whose information is this? If it belongs to a customer, a colleague or your company, it needs extra care.
- Is it public yet? Unreleased products, prices, results and plans should stay out.
- Can I describe it in general terms instead? Usually, yes. “A client in retail” works as well as the client’s name.
Your quick check before you press send
- No passwords, codes or PINs
- No ID, bank or card numbers
- No other people’s names, phone numbers or addresses
- No confidential work documents or client data
- No health details with a name attached
- Screenshots checked for anything in the corners
If everything on the list is clear, go ahead and get the help you need.
Common mistakes to avoid
- Pasting a whole email thread. Threads are full of names, numbers and signatures. Clean them first.
- Uploading screenshots without looking. A screenshot can show names, notifications or account details in the corner.
- Assuming “delete chat” erases everything. Deleting helps, but treat anything you’ve sent as possibly stored.
- Sharing other people’s stories. A friend’s health issue or a colleague’s money trouble isn’t yours to share, even to get advice.
Frequently asked questions
Is it safe to use ChatGPT, Gemini or Claude?
Yes, for most everyday tasks, as long as you keep private details out. The risk comes from what you type in, not from asking questions.
Do AI chatbots use my chats for training?
Some do by default, depending on the tool and your account type. Most let you turn this off in the settings. Work and business versions usually don’t train on your data, but check your tool’s policy.
Can I share work documents with AI?
Only if your workplace allows the tool you’re using, and only after removing private and confidential details. When in doubt, ask your manager or IT team first.
Is it okay to ask AI about my health or money?
Yes, for general questions, like “What does a high cholesterol number mean?” or “How does a credit score work?” Leave out your name, ID numbers and records, and talk to a doctor or financial adviser before making a big decision.
What should I do if I shared something by mistake?
Delete the chat, change any password or code you shared, and tell your manager or IT team if it was work information. Then check your settings so chats aren’t used for training.
The 5-second habit
Before you press send, ask: “Would I be okay if this ended up on a postcard?” If not, remove it or replace it with a placeholder. That one habit keeps you safe while you enjoy everything AI can do. Practice it for free in AI Essentials.
