WARNING: Never share these 5 things with ChatGPT, Gemini and other AI tools

NIN, BVN, bank details, medical results and photographs are among the information users should be cautious about uploading.

Users should remove or anonymise personal information that an AI tool does not need.

Confidential work documents should only be uploaded to AI tools approved by an employer.

You upload your bank statement to ChatGPT and ask it to explain where your money went.

Seconds later, it groups your spending, highlights your biggest expenses and suggests ways to save. Helpful, yes, but that screenshot may also contain your name, account number and transaction history.

The same thing can happen when you upload a medical report, photograph or document containing your NIN or BVN.

Now, the National Information Technology Development Agency (NITDA) is warning Nigerians to think carefully about what they share with AI platforms such as ChatGPT, Gemini, Claude, Copilot and Meta AI.

In its safe-use advisory, NITDA warned against sharing certain types of information with AI tools, as these platforms may retain, log, or use it to train models.

If sensitive information is later exposed through a breach or unauthorised access, it could potentially contribute to identity theft, impersonation or financial fraud.

So what should you keep out of your prompts?

FG seeks amendment to NIMC law to mandate foreigners to register for NIN. Your NIN and BVN are tied to your identity, while bank details can provide information about your finances.

There are plenty of situations where people might accidentally expose them. You may upload a completed application form because you want AI to check the wording.

You may send a bank statement because you want help analysing your spending. You may photograph a document without noticing that your identification number is visible in one corner.

If you want AI to analyse a bank statement, for example, remove your account number, name and other identifying details first. If you want it to check an application, cover your NIN, BVN, address and other information that is irrelevant to the task.

NITDA recommends removing, anonymising or pseudonymising personally identifiable information before using AI platforms. 

READ NEXT: Tinubu’s new NIMC Act: 7 things every Nigerian must do to avoid NIN problems

A medical report can contain much more than the result you want explained.

It may include your name, date of birth, hospital, test results, medications and information about a medical condition.

It is understandable that someone who receives a laboratory result they do not understand might turn to an AI chatbot and ask for an explanation. 

A chatbot may provide a useful general explanation, but it does not have the complete clinical picture and can produce inaccurate information.

If you want help understanding a report, remove your name and other identifying details first, and discuss medical decisions with a qualified professional.

Bar Beach Holiday ChatGPT Prompt Uploading a photograph to AI may seem harmless, particularly when you are simply asking the tool to improve the lighting, change the background or create a different version of the picture.

But photographs can affect more than the person in them. A picture might show your home, workplace, school, location or other people without their consent.

The risks attached to this include non-consensual intimate imagery, defamatory depictions, cyberbullying and exploitation, particularly involving children.

That does not mean an ordinary photograph uploaded for editing will necessarily be misused. It is simply worth considering whether the image needs to be uploaded at all, especially when other people are visible.

READ NEXT: American man uses ChatGPT to plan bank robbery, gets sentenced to 10 years in federal prison

Sensitive information does not always come in the form of a number or official document. You could reveal it while asking AI for help with a family problem.

For example, you might paste a message containing your partner’s full name and phone number and ask the chatbot to rewrite it. You might explain a problem involving your child and include their school, age and location.

The AI probably does not need those details to help you. Instead of giving a child’s name and school, you can simply say “my 12-year-old daughter”. Instead of pasting someone’s phone number, remove it before asking for help with the message.

You can still give the chatbot enough context to answer the question without identifying the people involved.

Do not share sensitive work documents with chatgpt and other AI tools. AI has also become a shortcut for everyday office tasks.

An employee might have a long report to summarise, an email to rewrite or a spreadsheet to analyse. Copying the material into an AI chatbot can seem like the quickest solution.

But work documents may contain customer information, financial records, internal plans, passwords, contracts or other material that an employee is not authorised to share with an external service.

Before uploading a work document, check whether your organisation permits the use of that AI service.

ALSO READ: Boko Haram now prompts AI? Report says terrorists used ChatGPT, Gemini and other AI tools to plan attacks

NITDA’s warning does not mean Nigerians should stop using AI. It is about being selective with what goes into the prompt.

Before uploading a document, photograph or screenshot, look at it carefully. If it contains information the AI does not need, remove it.

A chatbot does not need your NIN to rewrite an email. It does not need your BVN to help you create a budget. It does not need your full name and phone number to improve a message. And it does not need to know who your child is to help you think through a parenting problem.