What you will learn
- Decide what is safe to share and what to keep out
- Find and use the privacy controls in the app
- Avoid sharing passwords, keys, and personal data
- Handle sensitive work topics with care
- Spot outputs that should not be trusted or acted on
- Teach safe habits to people you work with
What not to paste
Some things should never go into a chatbot. Passwords, API keys, and access tokens are the clearest examples: once pasted, you should treat them as exposed and change them. The same caution applies to other people's personal data, such as full names with addresses, identification numbers, or medical details, and to confidential business information such as unreleased pricing, contracts, or customer lists.
A practical test is to ask whether the text would cause harm if it became public. If the answer is yes, do not paste it. Instead, replace sensitive details with made-up examples that keep the structure of the problem. You can usually get the same help with anonymised data as with real data, and the risk disappears.
Instead of pasting a real customer email, use a made-up version with the same structure: "Hi, I ordered item X on [date] and it arrived damaged. I would like a replacement." Ask ChatGPT to draft a reply to that version, then adapt the reply to the real email yourself.
Check the privacy controls in the app
ChatGPT's app includes privacy-related settings, and these change over time as features are added. Take a few minutes to open the settings and look for options about chat history, data use for training, and memory features. Read what each option says it does, in the app itself, rather than relying on summaries elsewhere, and choose the settings that match how careful you need to be.
If you use ChatGPT for work, check your employer's policy as well. Some organisations allow general use but forbid pasting company data, while others provide a business or team version with different data handling. Using a personal account for company work, or the reverse, can create problems that have nothing to do with the tool itself.
Be careful with sensitive topics
For health, legal, and financial questions, ChatGPT can explain general information, but it is not a doctor, a lawyer, or a financial adviser, and its answers are not personal advice. Use it to understand a topic and prepare questions for a professional, not to make a decision that affects your health, your rights, or your money. If an answer would change what you do in one of these areas, verify it with a qualified professional.
The same caution applies to content about children and to anything involving safety. If a situation is urgent or dangerous, contact the appropriate local service directly rather than asking a chatbot. ChatGPT is a text tool, not an emergency service.
Tip: Before pasting anything work-related, remove names, numbers, and identifiers and replace them with placeholders. If the answer still makes sense with placeholders, you never needed the real details in the first place.
Do not trust every output
ChatGPT can produce answers that sound authoritative and are wrong, including fake citations, incorrect figures, and confident explanations of things that do not exist. This is not a reason to avoid the tool, but it is a reason to verify anything that matters. Check facts against reliable sources, test code before using it, and be especially careful with quotes, statistics, and instructions for anything technical or regulated.
Be wary of outputs that ask you to take an irreversible action, such as deleting data, changing account settings, or sending money. Pause, check the instruction against an official source, and ask a knowledgeable person if you are unsure. A fluent explanation is not evidence that an instruction is safe.
Build safe habits for a team
If you introduce ChatGPT to a team, agree on simple ground rules first: what kinds of data are off limits, which account or plan to use for work, and how outputs should be checked before they are shared or published. Write the rules down in a short document and revisit them when features or policies change.
Encourage people to report mistakes they spot in outputs rather than hiding them. A team that shares examples of wrong or misleading answers learns faster than a team where everyone quietly works around the same problems alone. Safe use is a habit, not a setting, and habits are built by talking about real examples.
Frequently asked questions
Is it safe to paste personal information into ChatGPT?
Avoid pasting passwords, API keys, identification numbers, medical details, or other people's personal data. Replace sensitive details with made-up examples where possible, and check the current privacy settings in the app.
Can I stop my chats from being used for training?
The app includes privacy-related settings that change over time. Open the settings in the app itself and read what each option does, rather than relying on older summaries.
Can I use ChatGPT for work documents?
It depends on your employer's policy and the sensitivity of the documents. Many organisations allow general use but forbid pasting confidential company data. Check the policy first and anonymise details where possible.
Is ChatGPT's advice safe to follow for health or legal questions?
Treat it as general information, not personal advice. Use it to understand a topic and prepare questions, and verify important decisions with a qualified professional.
What should I do if ChatGPT gives a wrong or unsafe answer?
Do not act on it. Verify the claim against a reliable source, and for anything irreversible such as deleting data or sending money, check with an official source or a knowledgeable person first.