Is it safe? What happens to the things you type
Where your words go when you talk to an AI, what to keep to yourself, and simple habits that keep you in control.
This is the question careful people ask first — and asking it is a sign of good judgement, not of being behind the times. Here’s an honest answer, without scare tactics and without hand-waving.
Where do your words actually go?
When you type to a chatbot, your message travels over the internet to the company’s computers, which generate the reply. Your conversation is usually stored in your account so you can revisit it — like sent emails. Depending on the company and your settings, conversations may also be used to improve future versions of the AI.
So the honest summary is: a chatbot conversation is roughly as private as an email to a company — not a diary, not a locked drawer, but not a public noticeboard either.
The one rule that covers almost everything
Don’t type anything you’d be uncomfortable having a stranger at the company read.
That single habit protects you better than any setting. In practice it means keeping a few things out of the chat box entirely:
- Money numbers: bank account, card details, PIN codes, passwords.
- Identity numbers: national ID, passport, social security numbers.
- Other people’s private details: your neighbour’s health, your grandchild’s school — their stories aren’t yours to share with a company.
Everything else — recipe questions, letter drafting, “explain this word”, holiday planning — is comfortably within normal use.
“Can it empty my bank account?”
No. A chatbot on a website can only read what you type into it and write text back. It cannot reach into your computer, your files or your bank. The real money-related danger is older than AI: scammers. Which brings us to the part of this guide that matters most.
The real risk: people misusing AI on you
AI has made old tricks cheaper and more convincing. Three to know about:
- Cloned voices. A phone call that sounds exactly like a family member, urgently asking for money. If a call feels urgent and involves money, hang up and call the person back on the number you already have for them. Some families agree on a private code word for real emergencies — a surprisingly effective, low-tech defence.
- Polished scam messages. Fraud emails used to be full of odd grammar. AI writes them fluently now, so “it reads well” is no longer evidence of anything. Judge the request, not the spelling: anyone rushing you to pay, click, or share codes is suspect.
- Fake photos and videos. Seeing is no longer believing. If an image is shocking and pushes you to act (donate! share! panic!), let it wait a day — real news gets confirmed, fakes get debunked.
Notice the pattern: none of these require you to understand technology. They require the same scepticism you’d apply to a stranger at your door — plus one new habit: verify through a second channel you already trust.
Settings worth two minutes
Most chatbots have a settings menu (often behind your initials or a gear symbol). Two things are worth looking for once. On most chatbots, using your conversations for training is switched on unless you turn it off:
- An option to not use your conversations for training the AI.
- A button to delete conversations or your whole history.
Can’t find them? Ask the chatbot itself: “How do I stop my conversations being used for training?” It will point you to the right menu.
The bottom line
Using a chatbot is not risky the way handing out your card number is risky. Type freely about everyday things, keep the sensitive numbers out, and save your real vigilance for urgent requests involving money — wherever they come from. Next, the skill that separates wise users from everyone else: how to tell when AI is wrong.
