7 things you should never tell ChatGPT

Talking to ChatGPT can feel surprisingly similar to talking to another person. You ask a question, get an answer and continue the conversation. Before long, you might have shared where you live, where you work, what you are planning for the weekend and perhaps even what is worrying you.

But just because you can tell AI almost anything doesn’t mean you should.

So, what should you be careful about?

1. Passwords and banking details

This one is probably the easiest rule to follow: don’t type your passwords, PIN numbers, authentication codes or full bank card details into ChatGPT.

You can certainly ask for help creating a strong password or understanding how two-factor authentication works. But there is no reason to provide the real password you use for your email, online banking or social media.

The same applies to recovery codes and other information that could give someone access to one of your accounts.

2. Personal documents – at least not without checking them first

Uploading a document can be very useful. ChatGPT can help explain a contract, summarise a long report or make sense of a complicated letter.

The problem is that we often forget just how much information a single document can contain.

A passport, Citizen Card, bank statement, payslip or contract could include your full name, address, identification numbers, account information and even your signature. Before uploading anything, take a moment to look at the whole document and if certain personal details aren’t needed for your question, remove or cover them first.

3. Confidential information from work

Imagine you receive a long internal email and want ChatGPT to summarise it. Copy, paste, done.

But what else was in that email?

Perhaps a client’s name, financial information, an unpublished project, internal company figures or details about another employee.

AI can be incredibly useful at work, but it is worth treating confidential company information with the same care you would anywhere else. Before pasting an email, report or meeting notes into a chatbot, check what information you are actually sharing and whether your employer allows it.

4. “Tell me exactly what illness I have”

There is a big difference between using AI to better understand health information and asking it to diagnose you.

ChatGPT can explain medical terminology, help you prepare questions for your doctor or give general information about symptoms. But symptoms that sound similar can have very different causes, and an AI conversation cannot replace a proper medical assessment.

The same caution makes sense when the stakes are high in other areas, including legal and financial decisions.

Use AI to understand the subject better, not as the final word on what you should do.

5. “Make this major life decision for me”

Should I leave my job? Move abroad? End my relationship? Buy this house?

These are exactly the kinds of situations where ChatGPT can actually be useful. It can help you make a list of pros and cons, compare different scenarios or point out questions you haven’t considered.

What it cannot do is live with the consequences.

AI only knows the information you give it, and even a very detailed conversation cannot capture every part of your relationships, finances, priorities or circumstances.

Ask it to help you think through a decision. Be more careful about asking it to make the decision for you.

6. Information about other people that isn’t yours to share

Privacy isn’t only about protecting your own information.

It is easy to copy a WhatsApp conversation into ChatGPT and ask what someone “really meant”, or upload an email and ask for help replying. But those conversations may contain someone else’s phone number, address, personal problems, financial information or other details they never expected to be shared with an AI service.

The same goes for photographs and documents belonging to other people.

Before sharing something, it is worth asking a simple question: does ChatGPT actually need this information to answer me?

Usually, it doesn’t.

7. Anything you assume ChatGPT will automatically get right

Perhaps the most important thing not to give ChatGPT is complete trust.

AI can produce an answer that sounds extremely convincing and still get a name, date, statistic or source wrong. OpenAI itself warns that ChatGPT can sometimes provide incorrect or misleading information and recommends checking important information against reliable sources.

This becomes particularly important when you are using it for work, research, health, money or anything else where a mistake could matter.

Ask for sources. Open them. Check the figures. And if something sounds strange, don’t assume it must be true simply because the answer sounds confident.

What about privacy?

ChatGPT does give users some control over what happens to their conversations.

In Data Controls, users can choose whether their conversations can be used to improve OpenAI’s models. There is also a Temporary Chat option. Temporary conversations do not appear in chat history unless you choose to save them, do not create new memories and are not used to improve OpenAI’s models while they remain temporary. OpenAI says a copy may still be retained for up to 30 days for safety purposes.

But privacy settings don’t change the most useful rule of all: think before you share.

ChatGPT is at its best when it helps you research, organise, understand and create. You don’t need to give it every personal detail for it to be useful.

Set The Portugal News as your preferred source for news on Google

Add as a preferred source on Google