Picture a small business owner pasting a customer's email into a free ChatGPT to draft a reply. Name, address, what they ordered, the four thousand euro invoice, all of it, straight into the box. It feels harmless, and most of the time nothing bad happens. But whether it is harmless depends entirely on which chat window that text went into, and almost nobody checks before they hit enter.
The last post in this series told you to paste the real thing in, because real material beats describing it. That is still true. This post is the other half of that advice: there is a line, and crossing it is the one beginner mistake that can actually cost you. The good news is that staying on the right side of it takes about half a second once it becomes a habit.
Why This Is Worth Two Minutes
A short reality check, not a scare. When Cyberhaven looked at what people actually paste into ChatGPT, roughly eleven percent of it was confidential. Harmonic Security analyzed about a million prompts and found that eight and a half percent put sensitive data at risk, and more than half of those went into the free tier. The most quoted example is still Samsung: engineers pasted internal source code in to fix a bug, and the company banned the tool across the whole organization soon after.
None of that means AI is dangerous. It means the input box is not a private notebook. On free plans, your text can be used to train the model and can be seen by human reviewers. Once you know that, the rest is just common sense.
The Postcard Rule
Here is the single line to remember. On a free AI tool, treat anything you type like a postcard, not a sealed letter. Write what you would be fine with a stranger reading on its way through the post office. That one instinct catches almost every mistake before it happens, and you do not need to understand a word of how the model works to use it.
Three Buckets
Sort what you want to paste into three buckets and the decision gets easy.
Green, always fine. Anything already public or made up. Your published prices, your website copy, a general question, a draft where the real names are swapped for placeholders. Paste this freely, even on a free plan.
Yellow, fine if you anonymize. Real work material that has identifying bits in it. A customer email, a quote, a clause from a contract. This is exactly the stuff that makes answers good, so you do want to use it, you just strip the names and numbers first. The next section is the whole trick.
Red, never into a free tool. Anything you have a legal duty to protect or that would hurt if it leaked. Customer lists with personal data, health or financial records, passwords and access details, source code, anything covered by a confidentiality agreement. If you genuinely need AI on this kind of material, it goes into a paid business account, never a free one.
The One Trick: Anonymize, Don't Omit
This is where it all comes together. You do not have to choose between a good answer and keeping data safe. You replace the identifying parts with placeholders, let the AI do the work, then put the real details back in yourself afterwards.
Take the follow-up email from the previous post on writing prompts. Instead of "write a follow-up to Frau Berger about her four thousand euro patio quote," you write "write a follow-up to [CUSTOMER] about her [AMOUNT] [PROJECT] quote, warm tone, under a hundred words." The AI writes an equally good email. You paste her real name and the real number back in before you send. Thirty seconds, and nothing identifying ever actually left your screen. You get the quality the real material gives you, with none of the exposure.
Two Settings on Day One
Two small things, set once, and you have covered most of the risk.
First, turn off training and history in the settings of whatever tool you use. On most consumer tools there is a switch that stops your chats from being used to improve the model, and it is often on by default. Two minutes in the settings menu.
Second, if you handle real customer data regularly, pay for a Business or Team plan. On those, the provider contractually does not train on your input, and you can get the paperwork the law expects when someone else processes personal data for you. The free plan does not come with that paperwork, and that, not some vague feeling that "AI is risky," is the actual reason a free account is the wrong place for customer data. If you are still deciding which tool to put a paid plan on, the first post in this series walks through that choice.
The Mistakes Beginners Make
Uploading the whole document when you needed two paragraphs. The more you put in, the more there is to leak, and the model rarely needs the rest.
Pasting a screenshot with data still visible in a corner. The AI reads the entire image, not just the part you were thinking about.
Dropping in passwords, API keys, or logins to "let it help with the setup." Never. Those belong in a password manager, not a chat window.
Believing it forgets anyway. It does not, by default. Assume anything you type can be stored, and on a free plan, looked at by a person.
Make It a Habit, Not a Worry
The point of all this is not to make you nervous about a tool you just started to like. It is to turn one instinct into a reflex, the same way you already lower your voice when you say a customer's name out loud in a busy café. Anonymize the yellow stuff, keep a paid plan for the red stuff, and inside those lines paste as freely as you want. If you want the deeper, legal side of this, our guide to using AI in a GDPR-compliant way goes there. The next post in this series gets practical again: a plain list of the everyday jobs where this all starts paying off, the dozen tasks worth trying first thing Monday morning.
