AI Privacy Concerns: What Really Happens to What You Type
Whenever I show someone ChatGPT or Claude for the first time, there’s a moment — usually right before they type their first question — where they pause and ask: “Wait. Who sees this?”
Great instinct. Hold onto it.
I hold a cybersecurity certification through Google, though honestly, most of what I know about protecting people’s information came from 30-plus years of watching what goes wrong when we don’t. I use AI tools every day, and my answer isn’t “panic” or “it’s fine.” It’s the same advice I’d give about anything you say in public: be yourself, but watch what you hand over. Let me explain what’s actually going on.
Where does your typing go?
When you chat with an AI tool, your words don’t stay on your computer. They travel to the company’s servers — OpenAI’s, Anthropic’s, Google’s — where the AI does its thinking and sends an answer back.
Think of it like mailing a letter to a very fast pen pal. The pen pal is professional, and the envelope is sealed in transit. But the letter leaves your house. That’s the key mental shift.
What happens next depends on the company and your settings. Generally:
Your conversations are stored in your account history, like sent mail.
Some companies may use conversations to improve their AI — meaning humans might review samples, and your words could help train future versions. Many tools let you turn this off; it’s often on by default for free accounts.
Business and paid plans usually have stricter promises about not training on your data.
None of this means someone is reading your chats over morning coffee. It means your words become records that live on someone else’s computer. Treat them accordingly.
The never-paste list
Here’s the list I give every client. Print it, tape it near the monitor, no shame in that. Never put these into a chatbot:
Passwords or PINs. No tool needs them, ever. Anything asking for one is a scam, full stop.
Social Security numbers — yours or anyone else’s.
Full financial details. Card numbers, account numbers, routing numbers.
Medical records with names attached. Asking general health questions is fine. Pasting in a document with your name, birth date, and diagnosis is not.
Other people’s private information. Your customer list, your friend’s messy divorce details, an employee’s personnel file. It’s not yours to share with a robot.
Anything confidential from work. A client of mine once nearly pasted a whole unsigned business contract in for “a quick summary.” We paused, blanked out the names and dollar figures first, and then summarized. Same benefit, none of the exposure.
The rule of thumb: if you wouldn’t put it on a postcard, don’t put it in a chatbot. I think of it the way I run my kitchen: raw chicken gets its own cutting board, no exceptions, no matter how careful I’m feeling that day. Some things simply never touch the chat box. You can almost always get the AI’s help by describing your situation generally instead of pasting the raw document.
Settings worth checking (two minutes, tops)
Every major AI tool has privacy controls tucked into its settings. The names shift around, but look for:
”Improve the model for everyone” or “model training” — this is the switch that decides whether your chats can be used for training. Turn it off if that bothers you. In ChatGPT it’s under Settings, then Data Controls. Claude and Gemini have similar options.
Chat history controls. You can usually delete individual conversations or your whole history. Do a little cleanup now and then, like shredding old bank statements.
Temporary or incognito chat. Several tools now offer a mode where the conversation isn’t saved. Nice for anything you’d rather not have on the record.
Your account security itself. A strong, unique password and two-factor authentication protect your chat history the same way they protect your email. I’m an Apple guy — iPhone, Mac, iCloud — so I let iCloud Keychain generate and remember mine; on Windows or Android, any reputable password manager does the same job. Your AI account is only as private as its front door.
The calm takeaway
AI chatbots are not surveillance vans, and they’re not confessionals either. They’re services — useful, powerful, and run by companies with servers, policies, and settings.
Use them freely for the everyday stuff: drafting, learning, brainstorming, planning. Keep the crown jewels — passwords, identity numbers, other people’s secrets — out of the chat box. Flip the training switch if you want to. Then get on with enjoying the tools.
If you’d like someone to sit with you and check these settings on your own accounts — calmly, in plain English — I’d be glad to. Book a one-on-one session at indigosolutionsllc.com.