Is ChatGPT Safe for Seniors?
The honest answer is yes for ordinary use — but the risks people worry about are not the risks that actually exist. Here is what is worth your caution, and what is not.
Most people arrive at this question carrying a specific fear: that typing into ChatGPT will somehow expose their bank account, or let a stranger into their computer. That is not how it works, and it is worth clearing away first — because the fear stops people from using something genuinely useful, while leaving them unprotected against the thing that really does take money from older adults.
Can ChatGPT take my money or steal my identity?
A basic chat cannot see your bank account, email, or files by itself.
But ChatGPT can use apps and services you connect to it. Some can read information or take an action after you approve it.
Check what you connect. Read every confirmation before you approve it. And never type passwords, codes, or financial details into a chat.
Does ChatGPT save what I type?
Yes. Your conversations are stored in your account, and — this part surprises people — by default OpenAI may use consumer chats to help improve its models. That is the setting worth knowing about.
To stop your chats being used for training:
- Click your profile picture in the corner.
- Choose Settings.
- Choose Data Controls.
- Switch off Improve the model for everyone.
Your conversations still appear in your history — they simply stop being used to train the system. You can change it back at any time.
There is also a Temporary Chat option. These chats do not appear in your history or train the model. OpenAI may keep a copy for up to 30 days for safety.
What should I never type into ChatGPT?
A simple rule covers almost every case: treat the box like a letter that someone else might one day read. Most of what you would want to ask is perfectly fine. A short list is not.
- Passwords, of any kind, for any account.
- Verification codes — the six digits texted to you. No legitimate service ever needs these typed anywhere but its own login screen.
- Bank account, card, or routing numbers.
- Social Security or passport numbers.
- Full medical records, or anything you would not want associated with your name.
Asking it to explain a medical term, or to help you write a letter to your insurance company, is fine. Pasting in the whole document with your account number at the top is not — and usually is not necessary. Remove the numbers first; the answer will be just as good.
Is ChatGPT free, or will it charge me?
ChatGPT has a free version. It also has paid plans and shopping features.
Use chatgpt.com or the official app. Check the price and what you are buying before you enter card details.
What ChatGPT actually gets wrong
Here is the risk that is real and is underestimated: it can be completely wrong while sounding completely certain.
It does not know when it does not know. It will produce a confident paragraph containing an invented date, a law that does not exist, or a medication interaction that is simply untrue — in exactly the same calm, well-organized voice it uses when it is right. There is no tone change to warn you.
So the working rule is: use it to understand, not to decide. It is excellent at explaining what a term means, drafting a letter, or turning six pages into six sentences. It is not a doctor, a lawyer, or a financial adviser, and anything with money or health attached should be confirmed with an original source or a qualified person before you act on it.
A useful habit: when an answer matters, ask it "How do you know that? What should I check?" It will often tell you where it is uncertain.
The AI danger that actually costs older adults money
Almost none of the real harm comes from using ChatGPT. It comes from criminals using other AI tools against you — and the scale is not small.
The FBI's 2025 figures show adults aged 60 and over reported $7.7 billion in fraud losses, roughly a 60 percent rise on the previous year. Of the losses tied specifically to AI-enabled scams, $352 million came from people over 60.
The one to know about is voice cloning. A few seconds of someone's voice — from a voicemail, a video, a social media post — is now enough to generate convincing speech in that voice. The call comes late, the voice sounds like your grandchild, they are frightened, there has been an accident, and they need money immediately and beg you not to tell their parents.
The defense is simple, free, and takes one phone call to set up.
- Agree a family code word. Something no stranger could guess and that is not on the internet anywhere. Anyone calling in a genuine emergency can say it.
- Hang up and call back on the number you already have. Never the number they give you.
- Treat urgency itself as the warning sign. Real emergencies survive a five-minute pause. Scams do not — which is exactly why they insist you act now and tell no one.
- Never pay in gift cards, wire transfers, or cryptocurrency. No lawyer, hospital, or police force has ever asked for these.
Notice that none of these defenses require understanding how the technology works. They only require a habit.
How voice cloning actually works, how little audio it needs, and how to set up a family code word →
The short version
ChatGPT itself: safe for ordinary use. Connected apps may let it read information or take actions.
Watch what you type: no passwords, codes, account numbers, or ID numbers.
Turn off training in Settings → Data Controls if you would rather your chats were not used.
Trust it to explain, not to decide. It can be wrong in a confident voice.
The real threat is elsewhere: AI voice scams. Agree a family code word this week.
Being cautious about something new is not a failing. It is the correct instinct — it just needs to be pointed at the right thing.
The safest users are not the most careful ones.
This page gave you rules — do not type this, check that. Rules work until you meet a situation nobody wrote a rule for, and then they leave you guessing.
What holds up better is a rough picture of how the thing actually works. If you know why it invents a confident answer, you can feel in advance where the thin ice is, instead of trying to remember a list. That is the difference between being cautious and being hard to fool.
Getting Comfortable with AI is twenty cards that build exactly that picture — from "what is this thing" to "here is where it gets things wrong" — one clear idea and one useful thing to do on each. The whole set takes about 15 minutes to read — it is a set of cards, not a course. No jargon, no assumed background.
Free, and it stays free. 15 minutes — less time than it took to read this page.
Next: A respectful way to start using ChatGPT · What is a prompt? · Explore AI for Beginners · Return to Yilume
Sources. OpenAI, Data Controls FAQ, for the training setting and Temporary Chat retention. AARP's summary of the FBI Internet Crime Report for 2025, for fraud losses among adults 60 and over and the share attributed to AI-enabled scams. Figures were current as of August 2026; settings and prices change, so check them where it matters.